From jdohmen at uni-koblenz.de Thu Jan 9 07:26:05 2020
From: jdohmen at uni-koblenz.de (Joshua Dohmen)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Evaluation our curriculum with ergms?
Message-ID: <004201d5c701$1c4004f0$54c00ed0$@uni-koblenz.de>
Hello,
I hope you don't mind me asking my rookie question here. But the world of
network analysis is sometimes a bit confusing for me. I am trying to answer
a question concerning our study programme in pedagogy. I was wondering if
ERGMs or tERGMs are a possible way to answer it.
So what is it about? We have installed a format that we call ?study
partnership?. These are groups of students who meet once a week to discuss
and debate all study-related issues. We are now wondering whether the study
partnerships really have an influence on the exchange of knowledge between
students. So I collected network data at two different time points to check
whether the importance of study partnerships for knowledge exchange is
increasing or decreasing.
Is it possible to answer such a question in principle with tERGMs? If so, I
would try to get my data into statnet using something like
nodematch('studypartnerships')? The results should give me a tie-formation
and a tie-dissolution probability which is not quite clear to me right now.
Thank you for any advice on what to look out for! Sorry for my lack of
knowledge on this. The tutorials on github are fantastic. I really want to
learn more about these methods.
Kind regards,
Joshua
--
Dipl.-P?d. Joshua Dohmen
Universit?t Koblenz-Landau, Campus Koblenz, Fachbereich 1:
Bildungswissenschaften, Institut f?r P?dagogik, Abteilung P?dagogik
Arbeitsbereich: Forschung und Entwicklung in Organisationen
Tel: 0049/261/287-1894
R: E 224
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ssk170 at scarletmail.rutgers.edu Sun Jan 12 01:37:53 2020
From: ssk170 at scarletmail.rutgers.edu (Sergei Kostiaev)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] creating nodal attribute vector from edge attribute
vector
Message-ID:
hello,
i've created dataset (see the file attached) which has 3 variables:
organizations that take positions on an issue (1 mode), letters that
express organizations' views (2 mode), position that organization takes (-1
opposition, 1 support)
that third variable is an edge attribute
any suggestions how i can create node attribute from edge attribute?
Best regards,
Dr. Sergei Kostiaev
PhD Candidate
Edward J. Bloustein School of Planning and Public Policy;
Part-Time Lecturer
Political Science Department
Rutgers University;
U.S. cell +1(202)361-7672
http://bloustein.rutgers.edu/kostiaev/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coalitionsAHPs.Rds
Type: application/octet-stream
Size: 18734 bytes
Desc: not available
URL:
From morrism at uw.edu Thu Jan 16 07:00:23 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Evaluation our curriculum with ergms?
In-Reply-To: <004201d5c701$1c4004f0$54c00ed0$@uni-koblenz.de>
References: <004201d5c701$1c4004f0$54c00ed0$@uni-koblenz.de>
Message-ID:
> So what is it about? We have installed a format that we call ?study partnership?. These are
> groups of students who meet once a week to discuss and debate all study-related issues. We are
> now wondering whether the study partnerships really have an influence on the exchange of
> knowledge between students. So I collected network data at two different time points to check
> whether the importance of study partnerships for knowledge exchange is increasing or decreasing.
>
> Is it possible to answer such a question in principle with tERGMs? If so, I would try to get my
> data into statnet using something like nodematch('studypartnerships')? The results should give
> me a tie-formation and a tie-dissolution probability which is not quite clear to me right now.
Thanks for your email. I think it would help to get a bit more clarity
on what your research question is, and for that, it would help to be
specific about
1. how you define a "link" between two students (conceptually -- what are
you trying to measure)
2. how did you operationalize this in the survey? (i.e., what are the
actual questions you asked each student)
3. did you collect data from every student in the study groups?
Then we can tell what kind of process you're trying to understand, and
whether that maps to what tERGMs can do.
best,
mm
>
> --
>
> Dipl.-P?d. Joshua Dohmen
>
> Universit?t Koblenz-Landau, Campus Koblenz, Fachbereich 1: Bildungswissenschaften, Institut f?r
> P?dagogik, Abteilung P?dagogik
>
> Arbeitsbereich: Forschung und Entwicklung in Organisationen
>
> Tel: 0049/261/287-1894
>
> R: E 224
>
>
>
>
>
>
>
****************************************************************
Professor of Sociology and Statistics
Box 354322
University of Washington
Seattle, WA 98195-4322
Office: (206) 685-3402
Dept Office: (206) 543-5882, 543-7237
Fax: (206) 685-7419
morrism@u.washington.edu
http://faculty.washington.edu/morrism/
From morrism at uw.edu Thu Jan 16 07:07:05 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] creating nodal attribute vector from edge
attribute vector
In-Reply-To:
References:
Message-ID:
Hi Sergei,
It's not quite clear what you are asking here.
First, it sounds like you have a 2 mode network, orgs x issues, with
letters (+/-) as the ties. Is that right?
What kind of node attributes are you thinking about? Are you trying to
represent the node position in this network? Or calculate some kind of
egocentric index based on the number/type of ties?
best,
mm
On Sun, 12 Jan 2020, Sergei Kostiaev wrote:
> hello,
> i've?created dataset (see the file attached) which has 3 variables: organizations that take
> positions on an issue (1 mode), letters that express organizations' views (2 mode), position
> that organization takes (-1 opposition, 1 support)
> that third variable is an edge attribute
>
> any suggestions how i can create node attribute from edge attribute?
>
>
>
>
>
> Best regards,
> Dr. Sergei KostiaevPhD Candidate
> Edward J. Bloustein School of Planning and Public Policy;
> Part-Time Lecturer
> Political Science Department
> Rutgers University;
> U.S. cell +1(202)361-7672
> http://bloustein.rutgers.edu/kostiaev/
>
>
>
>
>
****************************************************************
Professor of Sociology and Statistics
Box 354322
University of Washington
Seattle, WA 98195-4322
Office: (206) 685-3402
Dept Office: (206) 543-5882, 543-7237
Fax: (206) 685-7419
morrism@u.washington.edu
http://faculty.washington.edu/morrism/
From jdohmen at uni-koblenz.de Fri Jan 17 02:15:52 2020
From: jdohmen at uni-koblenz.de (Joshua Dohmen)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Evaluation our curriculum with ergms?
In-Reply-To:
References: <004201d5c701$1c4004f0$54c00ed0$@uni-koblenz.de>
Message-ID: <004d01d5cd1f$18dec2f0$4a9c48d0$@uni-koblenz.de>
Hello,
thank you for the questions. I don't want to take too much of your time, so I'll try to keep it as short as possible without leaving out anything important. I use a pedagogical theory that was written in German, so there might be some translation issues.
In general I am interested in learning processes within organizations. Here the exchange of "organization-related knowledge" plays an important role. In my case: When students join the university, they not only have to learn the contents of their curriculum, but also how the university works as an organizational system. Based on this assumption, we have implemented study partnerships in our BA Pedagogy program to encourage the knowledge sharing. My question is whether the study partnerships are of lasting importance or only have an effect in the initial phase of the study.
More detailed to the questions:
1. how you define a "link" between two students? Since I cannot look into the heads of my students, I have to measure knowledge sharing indirectly. So I ask who is usually present at certain typical occasions for the exchange of organization-related knowledge (assuming that it really happens).
2. how did you operationalize this in the survey? Together with the students, I have developed four different name generators for (informal and formal) occasions of knowledge sharing, e.g.: "If discussions about the university are held in the cafeteria during lunch break, which fellow students from your program and semester are usually involved in these discussions?"
3. did you collect data from every student in the study groups? Yes, I created a web based form with an auto-complete function ("type the first few letters of a name") and automatic name validation ("this name does not exist"). Nearly all students of the year took part in the survey. Of course there were some dropouts between the two survey dates (June 2018 and Nov 2019) because some students left the university. For the analysis the names of the students were transformed into unique tokens ("Joshua Dohmen" --> "txQ1zgP23"), so that I can't link the results (e.g. indegree) to a certain clear name.
Maybe important: The assignment of the students to the study partnership was random.
This is what the graphs look like: https://userpages.uni-koblenz.de/~organisation/Network.png
Many thanks for all your help!
Joshua
-----Urspr?ngliche Nachricht-----
Von: martina morris
Gesendet: Donnerstag, 16. Januar 2020 16:00
An: Joshua Dohmen
Cc:
Betreff: Re: [statnet_help] Evaluation our curriculum with ergms?
> So what is it about? We have installed a format that we call ?study
> partnership?. These are groups of students who meet once a week to
> discuss and debate all study-related issues. We are now wondering
> whether the study partnerships really have an influence on the
> exchange of knowledge between students. So I collected network data at two different time points to check whether the importance of study partnerships for knowledge exchange is increasing or decreasing.
>
> Is it possible to answer such a question in principle with tERGMs? If
> so, I would try to get my data into statnet using something like
> nodematch('studypartnerships')? The results should give me a tie-formation and a tie-dissolution probability which is not quite clear to me right now.
Thanks for your email. I think it would help to get a bit more clarity on what your research question is, and for that, it would help to be specific about
1. how you define a "link" between two students (conceptually -- what are you trying to measure)
2. how did you operationalize this in the survey? (i.e., what are the actual questions you asked each student)
3. did you collect data from every student in the study groups?
Then we can tell what kind of process you're trying to understand, and whether that maps to what tERGMs can do.
best,
mm
>
> --
>
> Dipl.-P?d. Joshua Dohmen
>
> Universit?t Koblenz-Landau, Campus Koblenz, Fachbereich 1:
> Bildungswissenschaften, Institut f?r P?dagogik, Abteilung P?dagogik
>
> Arbeitsbereich: Forschung und Entwicklung in Organisationen
>
> Tel: 0049/261/287-1894
>
> R: E 224
>
>
>
>
>
>
>
****************************************************************
Professor of Sociology and Statistics
Box 354322
University of Washington
Seattle, WA 98195-4322
Office: (206) 685-3402
Dept Office: (206) 543-5882, 543-7237
Fax: (206) 685-7419
morrism@u.washington.edu
http://faculty.washington.edu/morrism/
From jdohmen at uni-koblenz.de Wed Feb 5 01:14:55 2020
From: jdohmen at uni-koblenz.de (Joshua Dohmen)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] STERGM: links driving attribute change
Message-ID: <006c01d5dc04$bb511210$31f33630$@uni-koblenz.de>
Hi list,
Is it possible to model weather tie-formation/tie-dissolution leads to
attribute change (using stergm)?
This been a topic here:
[statnet_help] STERGM and selection vs influence:
http://mailman13.u.washington.edu/mailman/htdig/statnet_help/2014/001713.htm
l
[statnet_help] STERGM with time-varying nodal covariates:
http://mailman13.u.washington.edu/pipermail/statnet_help/2012/001324.html
What would the interpretation look like?
If a node i is involved in tie-formation (between t1 and t2) -- ... is it
likely for i to take over the attribute value of j? .is it likely for i and
j to have the same attribute value at t2?
And when the attribute is numeric -- . the value difference is getting
smaller?
The postings are already a little bit older. If there are some new
developments that I might not have found, I would be happy to hear about it.
Regards,
Joshua
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From morrism at uw.edu Wed Feb 5 11:14:38 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] STERGM: links driving attribute change
In-Reply-To: <006c01d5dc04$bb511210$31f33630$@uni-koblenz.de>
References: <006c01d5dc04$bb511210$31f33630$@uni-koblenz.de>
Message-ID:
Hi Josh,
It is not possible to do that using STERGM alone -- STERGM, like ERGM,
is a model for tie process only (not nodal attributes).
But you could probably use the EpiModel package to do this. EpiModel uses
ergm/tergm to represent the tie formation/dissolution process, but then
uses other code to update nodal attributes as a function of the ties. It
was designed for epidemic modeling, where ties allow for transmission (and
thus changing nodal status) over time.
We have lots of info up online, and there's a dedicated EpiModel help
list. See: epimodel.org and http://statnet.github.io/nme/
I don't know if this will meet your needs, but let us know if you have any
questions.
best,
mm
On Wed, 5 Feb 2020, Joshua Dohmen wrote:
>
> Hi list,
>
>
>
> Is it possible to model weather tie-formation/tie-dissolution leads to attribute change (using
> stergm)?
>
>
>
> This been a topic here:
>
> [statnet_help] STERGM and selection vs influence:
> http://mailman13.u.washington.edu/mailman/htdig/statnet_help/2014/001713.html
>
> [statnet_help] STERGM with time-varying nodal covariates:
> http://mailman13.u.washington.edu/pipermail/statnet_help/2012/001324.html
>
> What would the interpretation look like?
>
> If a node i is involved in tie-formation (between t1 and t2) -- ... is it likely for i to take
> over the attribute value of j? ?is it likely for i and j to have the same attribute value at t2?
>
> And when the attribute is numeric -- ? the value difference is getting smaller?
>
>
>
> The postings are already a little bit older. If there are some new developments that I might not
> have found, I would be happy to hear about it.
>
>
>
> Regards,
>
> Joshua
>
>
>
>
>
****************************************************************
Professor of Sociology and Statistics
Box 354322
University of Washington
Seattle, WA 98195-4322
Office: (206) 685-3402
Dept Office: (206) 543-5882, 543-7237
Fax: (206) 685-7419
morrism@u.washington.edu
http://faculty.washington.edu/morrism/
From sefutch at ncsu.edu Thu Feb 13 07:55:28 2020
From: sefutch at ncsu.edu (Sara Futch)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Bipartite ERGM terms?
Message-ID:
Hi all,
Does anyone have much experience with Bipartite ERGMs in statnet? I'm
having a hard time finding the right terms.
I want to do an analysis similar to Stephens, Chen, and Butler (2016)
to get at
activity spread and popularity spread in my network. I have quite a large
network (3651 x 629); I believe this is too large for PNet, which is what
these authors used. Has anyone done a similar analysis in statnet? I think
that GWESP seems on track with what I want, but it's only for one-mode
networks.
I appreciate any insight, thank y'all!
Sara Futch
--
Graduate Research Assistant
Fisheries, Wildlife, and Conservation Biology
Human Dimensions Minor
College of Natural Resources
North Carolina State University
she | her | hers
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From filippo.santi at unifi.it Mon Feb 17 01:32:08 2020
From: filippo.santi at unifi.it (Filippo Santi)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Fwd: help
In-Reply-To:
References:
Message-ID:
Dear all,
I am running an ERGM model on world level bilateral trade flows, using a
few additional covariates at node as well as at dyad level. I also add the
GWESP term.
I have several questions related to mcmc.diagnostic(), which I launch to
check for degeneracy. I am sure that they might turn out to be very simple
those who know the principle of MCMC, but unfortunately, it is not my case,
and I am a bit stuck for what concerns the interpretation.
In particular:
1) what do the axes refer to?
2) How should I interpret the fact that my estimates are not fluctuating
around zero, but still relatively stable around a certain value?
3) My distribution is skewed (often to the left) and this is reflected also
in the left hand side plot, with the initialised value lying below the
central value. What is the exact meaning of this? Where should I look to
try to solve this?
4) To make things faster, I used multiple cores to perform the estimation.
The diagnostic of some of the cores is often completely shifted with
respect to the other. Is it normal? if it is not, what could explain that?
I report below the model. The MCMC.DIAGNOSTIC() plots are instead attached
as PNG.
Thank you in advance for any help.
--
*Filippo Santi*
*Phd in Development Economics*
*-----------------------------------------------------*
*grav.model.trade.12 <- ergm(net.trade.2010.redux ~ edges + mutual
+ edgecov(net.migration.2010, 'weight')
+ nodeicov('gdpcap') + nodeocov('gdpcap')
+ edgecov(net.distance.2010, 'log_dist') +
edgecov(net.colony.2010, 'colony') +
edgecov(net.contiguity.2010, 'contig') +
edgecov(net.language.2010, 'comlang_ethno') +
gwesp(decay = 0.25, fixed = TRUE) # Didier et al. 2019: decay = 0.25
standard in literature... , eval.loglik = T,
control=control.ergm(seed=1,MCMC.burnin = 50000,MCMC.interval = 3000,
MCMC.samplesize = 70000, MCMLE.maxit = 10000,
main.method = "Stepping",
parallel=np,
parallel.type
="PSOCK"))summary(grav.model.trade.12)==========================Summary of
model fit==========================Formula: net.trade.2010.redux ~ edges
+ mutual + edgecov(net.migration.2010, "weight") + nodeicov("gdpcap") +
nodeocov("gdpcap") + edgecov(net.distance.2010, "log_dist") +
edgecov(net.colony.2010, "colony") + edgecov(net.contiguity.2010,
"contig") + edgecov(net.language.2010, "comlang_ethno") + gwesp(decay =
0.25, fixed = TRUE)Iterations: NA Stepping MLE Results:
Estimate Std. Error MCMC % z value Pr(>|z|) edges
-6.675e+00 8.859e-12 0 -7.534e+11 < 1e-04 ***mutual
2.584e+00 3.346e-12 0 7.722e+11 < 1e-04 ***edgecov.weight
6.874e-07 1.842e-07 0 3.731e+00 0.000191 ***nodeicov.gdpcap
6.671e-06 1.057e-06 0 6.313e+00 < 1e-04 ***nodeocov.gdpcap
2.436e-06 1.081e-06 0 2.253e+00 0.024287 * edgecov.log_dist
-3.065e-01 7.275e-11 0 -4.213e+09 < 1e-04 ***edgecov.colony
6.794e-01 3.086e-13 0 2.202e+12 < 1e-04 ***edgecov.contig
8.230e-01 5.504e-13 0 1.495e+12 < 1e-04 ***edgecov.comlang_ethno
2.955e-01 1.770e-12 0 1.670e+11 < 1e-04 ***gwesp.fixed.0.25
4.617e+00 1.125e-11 0 4.105e+11 < 1e-04 ***---Signif. codes: 0
?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1 Null Deviance: 36606 on
26406 degrees of freedom Residual Deviance: 11473 on 26396 degrees of
freedom AIC: 11493 BIC: 11575 (Smaller is better.)#
-------------------------------------------------------------------------*
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: degeneracy_3_12feb2020.png
Type: image/png
Size: 10575 bytes
Desc: not available
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: degeneracy_2_12feb2020.png
Type: image/png
Size: 11540 bytes
Desc: not available
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: degeneracy_12feb2020.png
Type: image/png
Size: 5276 bytes
Desc: not available
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: degeneracy_4_12feb2020.png
Type: image/png
Size: 10352 bytes
Desc: not available
URL:
From goodreau at uw.edu Mon Feb 17 19:15:44 2020
From: goodreau at uw.edu (Steven Goodreau)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Network Modeling for Epidemics -- Summer short
course at the University of Washington -- 17-21 August, 2020
In-Reply-To:
References:
Message-ID: <17c3ec23-f776-bdd7-ba4d-efdb41c6f856@uw.edu>
############################################################################
Network Modeling for Epidemics
Summer short course at the University of Washington
17-21 August, 2020
Network Modeling for Epidemics (NME) is a 5-day short course at the
University of Washington that provides an introduction to stochastic
network models for infectious disease transmission dynamics, with a
focus on empirically based modeling of HIV transmission. It is a
''hands-on'' course, using the EpiModel software package in R
(www.epimodel.org). EpiModel provides a unified framework for
statistically based modeling of dynamic networks from empirical data,
and simulation of epidemic dynamics on these networks. It has a flexible
open-source platform for learning and building several types of epidemic
models: deterministic compartmental, stochastic individual-based, and
stochastic network models. Resources include simple models that run in a
browser window, built-in generic models that provide basic control over
population contact patterns, pathogen properties and demographics, and
templates for user-programmed modules that allow EpiModel to be extended
to the full range of pathogens, hosts, and disease dynamics for advanced
research. This course will touch on the deterministic and
individual-based models, but its primary focus is on the theory, methods
and application of network models.
The course uses a mix of lectures, tutorials, and labs with students
working in small groups. On the final day, students work to develop an
EpiModel prototype model (either individually or in groups based on
shared research interests), with input from the instructors, including
the lead EpiModel software developer, Dr. Samuel Jenness.
Returning students: We encourage previous attendees with active modeling
projects to apply to return for a refresher course. The EpiModel package
has been significantly enhanced over the last few years. Returning
students with active projects will have the opportunity to work with
course instructors to address key challenges in the design of their
network model code.
*Dates and location:*
The course will be taught from Monday, August 17 to Friday, August 21 on
the University of Washington campus in Seattle.
*Costs:*
Course fee is $800. Travel and accommodation costs are the
responsibility of the participant, although discounted hotel rates are
available. We offer a limited number of fee waivers for pre-doctoral
students or for attendees from low income countries. These cover waiver
of the registration fee only; travel and accommodation are still the
responsibility of the fee waiver recipient.
*Application dates and decision dates:*
*??? Apr 19: Fee-waiver application deadline. Decisions will be made by
Apr 30, and response required by May 15.
*??? May 17:? General application deadline. Decisions will be made by
May 31 and a response required by June 15.
*??? A waitlist will be established along with rolling admission through
June 30 if space allows.
*Application:*
Apply online at https://catalyst.uw.edu/webq/survey/morrism/385149
Course website and more information: http://statnet.github.io/nme
Please free to share widely!
Note: We will also be teaching this course in Antwerp, Belgium, 7-11
September.? More information can be found here:
https://www.uantwerpen.be/en/summer-schools/modelling-infectious-diseases/
Yours,
Martina Morris, Steve Goodreau and Samuel Jenness
--
*****************************************************************
Steven M. Goodreau / Professor / Dept. of Anthropology
Physical address: Denny Hall M236
Mailing address: Campus Box 353100 / 4216 Memorial Way NE
Univ. of Washington / Seattle WA 98195
1-206-685-3870 (phone) /1-206-543-3285 (fax)
http://faculty.washington.edu/goodreau
*****************************************************************
From Darren.Gillis at umanitoba.ca Fri Feb 21 07:13:17 2020
From: Darren.Gillis at umanitoba.ca (Darren Gillis)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Likelihood and AIC from mtergm() in the btergm
package by Leifeld
Message-ID:
Hello,
I am not sure if this is the correct place to post regarding btergm, but I thought I'd try as other related posts appear in the archives.
I am currently working with a dynamic network (four annual cross-sections) of vessel associations in a North Sea trawl fishery. I am exploring both tergm::stergm() and btergm::mtergm() as alternative modeling methods. I am interested in using AIC to compare alternative models within (not between) each of these packages (in addition to examining MCMC chains, checking degeneracy and GOF). I can easily extract these values from stergm(), but in mtergm() the S4 slots for likelihood, aic, and bic mtergm are all NA. In addition, attempts to directly obtain these values with "AIC(mtergm.model) " result in:
Error in UseMethod("logLik") :
no applicable method for 'logLik' applied to an object of class "mtergm"
Are these slots place holders for compatibility with other packages or have I missed something in the manual and associated publication (Leifeld et al. 2018)? I think the answer is pretty clear, but I wanted to double check before I moved on. Thank you for considering my query.
Dr. Darren M. Gillis
Professor, Biological Sciences
Faculty of Science
University of Manitoba
?Ignorance more frequently begets confidence than does knowledge.?
? Charles Darwin
"..., reality must take precedence over public relations, for Nature cannot be fooled."
? Richard P. Feynman
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From philip.leifeld at essex.ac.uk Sat Feb 22 08:04:25 2020
From: philip.leifeld at essex.ac.uk (Leifeld, Philip)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Likelihood and AIC from mtergm() in the btergm
package by Leifeld
In-Reply-To: <9f4bf88a7c7c4165a91383c7813237b9@LNXP265MB0714.GBRP265.PROD.OUTLOOK.COM>
References: <9f4bf88a7c7c4165a91383c7813237b9@LNXP265MB0714.GBRP265.PROD.OUTLOOK.COM>
Message-ID:
Hi Darren,
While the btergm function implements a bootrapping correction to MPLE
standard errors in a TERGM (i.e., for a panel of many networks), mtergm
is just a convenience wrapper for the ergm::ergm function with MCMLE
(i.e., for few waves) based on a block-diagonal matrix with an offset
term. The original ergm object is stored inside the output, and you can
access it via mtergm.model@ergm. However, the ergm package does not seem
to define a working logLik method for models with an offset term, hence
logLik(mtergm.model@ergm) does not seem to work. Upon inspecting a
fitted model, I can see that a log likelihood value is stored in the
$loglikelihood slot. Provided that this is indeed the log likelihood,
which the statnet developers should be able to confirm, I suppose you
may be able to calculate the desired quantities as follows:
ll <- as.numeric(mtergm.model@ergm$loglikelihood)
aic <- -2 * ll + (2 * length(mtergm.model@ergm$coef))
bic <- -2 * ll + (2 * log(nobs(mtergm.model@ergm)))
That said, I am not sure how well-defined BIC is in a network context
given its reliance on the number of observations, which is perhaps
ambiguous in a network context.
If you have any follow-up questions, please feel free to email me
directly or add an entry to the issue tracker for the btergm package on
GitHub: https://github.com/leifeld/btergm/issues
Best regards,
Philip
--
Professor of Comparative Politics
Department of Government
University of Essex
http://www.philipleifeld.com
On 21/02/2020 15:13, Darren Gillis wrote:
> Hello,
>
>
> I am not sure if this is the correct place to post regarding btergm, but
> I thought I'd try as other related posts appear in the archives.
>
>
> I am currently working with a dynamic network (four annual
> cross-sections) of vessel associations in a North Sea trawl fishery. I
> am exploring both tergm::stergm() and btergm::mtergm() as alternative
> modeling methods. I am interested in using AIC to compare alternative
> models within (not between)?each of these packages (in addition to
> examining? MCMC chains, checking?degeneracy and GOF). I can easily
> extract these values from stergm(), but in mtergm()?the S4?slots for
> likelihood, aic, and bic?mtergm are all NA. In addition, attempts to
> directly obtain these values with "AIC(mtergm.model) "?result in:
>
>
> Error in UseMethod("logLik") : no applicable method for 'logLik' applied
> to an object of class "mtergm"
>
>
>
> Are these slots place holders for compatibility with other packages or
> have I missed?something in the manual and associated publication
> (Leifeld et al. 2018)?? ?I think the answer is pretty clear, but I
> wanted to double check before I moved on.?Thank you for considering my
> query.
>
>
> Dr. Darren M. Gillis
> Professor,?Biological Sciences
> Faculty of Science
> University of Manitoba
>
> ?Ignorance more frequently begets confidence than does knowledge.?
> ? Charles Darwin
>
>
> "...,? reality must take precedence over public relations, for Nature
> cannot be fooled."
> ? Richard P. Feynman
>
>
>
From Darren.Gillis at umanitoba.ca Sat Feb 22 09:06:23 2020
From: Darren.Gillis at umanitoba.ca (Darren Gillis)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Likelihood and AIC from mtergm() in the btergm
package by Leifeld
In-Reply-To:
References: <9f4bf88a7c7c4165a91383c7813237b9@LNXP265MB0714.GBRP265.PROD.OUTLOOK.COM>,
Message-ID: <8e20da3c1c7e40ec8fb7899c32e082ea@umanitoba.ca>
Hi Philip,
Thank you for your rapid reply! This is very helpful.
First, I apologize for stating that the likelihood, aic, and bic slots contain NA when in fact they contain NaN. The same issue for me, but it could be important (or misleading) for someone trying to determine the root of the issue.
Especially, I want to thank you also for the quick AIC hack. It will serve my purposes for now, but I will wait for input from the statnet team on the meaning of ergm$likelihood before I use these values in my manuscript. And I always have the auc.pr values across the panel of networks. In fact, it may be just as good for my goal, but AIC is more familiar for most fisheries reviewers.
I am exploring both your package and STERGMs for this dynamic network, and I greatly appreciate the work that you and the statnet team have placed in your packages.
Warmest regards, Darren
Dr. Darren M. Gillis
Professor, Biological Sciences
Faculty of Science
University of Manitoba
________________________________
From: Leifeld, Philip
Sent: February 22, 2020 10:04 AM
To: statnet_help@u.washington.edu
Cc: Darren Gillis
Subject: Re: [statnet_help] Likelihood and AIC from mtergm() in the btergm package by Leifeld
Hi Darren,
While the btergm function implements a bootrapping correction to MPLE
standard errors in a TERGM (i.e., for a panel of many networks), mtergm
is just a convenience wrapper for the ergm::ergm function with MCMLE
(i.e., for few waves) based on a block-diagonal matrix with an offset
term. The original ergm object is stored inside the output, and you can
access it via mtergm.model@ergm. However, the ergm package does not seem
to define a working logLik method for models with an offset term, hence
logLik(mtergm.model@ergm) does not seem to work. Upon inspecting a
fitted model, I can see that a log likelihood value is stored in the
$loglikelihood slot. Provided that this is indeed the log likelihood,
which the statnet developers should be able to confirm, I suppose you
may be able to calculate the desired quantities as follows:
ll <- as.numeric(mtergm.model@ergm$loglikelihood)
aic <- -2 * ll + (2 * length(mtergm.model@ergm$coef))
bic <- -2 * ll + (2 * log(nobs(mtergm.model@ergm)))
That said, I am not sure how well-defined BIC is in a network context
given its reliance on the number of observations, which is perhaps
ambiguous in a network context.
If you have any follow-up questions, please feel free to email me
directly or add an entry to the issue tracker for the btergm package on
GitHub: https://github.com/leifeld/btergm/issues
Best regards,
Philip
--
Professor of Comparative Politics
Department of Government
University of Essex
http://www.philipleifeld.com
On 21/02/2020 15:13, Darren Gillis wrote:
> Hello,
>
>
> I am not sure if this is the correct place to post regarding btergm, but
> I thought I'd try as other related posts appear in the archives.
>
>
> I am currently working with a dynamic network (four annual
> cross-sections) of vessel associations in a North Sea trawl fishery. I
> am exploring both tergm::stergm() and btergm::mtergm() as alternative
> modeling methods. I am interested in using AIC to compare alternative
> models within (not between) each of these packages (in addition to
> examining MCMC chains, checking degeneracy and GOF). I can easily
> extract these values from stergm(), but in mtergm() the S4 slots for
> likelihood, aic, and bic mtergm are all NA. In addition, attempts to
> directly obtain these values with "AIC(mtergm.model) " result in:
>
>
> Error in UseMethod("logLik") : no applicable method for 'logLik' applied
> to an object of class "mtergm"
>
>
>
> Are these slots place holders for compatibility with other packages or
> have I missed something in the manual and associated publication
> (Leifeld et al. 2018)? I think the answer is pretty clear, but I
> wanted to double check before I moved on. Thank you for considering my
> query.
>
>
> Dr. Darren M. Gillis
> Professor, Biological Sciences
> Faculty of Science
> University of Manitoba
>
> ?Ignorance more frequently begets confidence than does knowledge.?
> ? Charles Darwin
>
>
> "..., reality must take precedence over public relations, for Nature
> cannot be fooled."
> ? Richard P. Feynman
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From Darren.Gillis at umanitoba.ca Wed Feb 26 18:09:04 2020
From: Darren.Gillis at umanitoba.ca (Darren Gillis)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Likelihood and AIC from mtergm() in the btergm
package by Leifeld
In-Reply-To: <8e20da3c1c7e40ec8fb7899c32e082ea@umanitoba.ca>
References: <9f4bf88a7c7c4165a91383c7813237b9@LNXP265MB0714.GBRP265.PROD.OUTLOOK.COM>,
,
<8e20da3c1c7e40ec8fb7899c32e082ea@umanitoba.ca>
Message-ID: <757fbf3cb56b49d5adb377b2d313fb21@umanitoba.ca>
After digging into the GIT code (which I should have done before) I see that loglikelihood value returned by ergm() is:
"The approximate change in log-likelihood in the last iteration. The value is only approximate because it is estimated based on the MCMC random sample."
and not the model value that I require. I was wondering if the mle.lik value is available for "a block-diagonal matrix with an offset term" described by Dr. Leifeld below. If not , I will continue with AUC based GOF statistics. Thank you for taking the time to consider my query.
Regards, Darren
Dr. Darren M. Gillis
Professor, Biological Sciences
Faculty of Science
University of Manitoba
________________________________
From: Darren Gillis
Sent: February 22, 2020 11:06 AM
To: Leifeld, Philip; statnet_help@u.washington.edu
Cc: morrism@u.washington.edu
Subject: Re: [statnet_help] Likelihood and AIC from mtergm() in the btergm package by Leifeld
Hi Philip,
Thank you for your rapid reply! This is very helpful.
First, I apologize for stating that the likelihood, aic, and bic slots contain NA when in fact they contain NaN. The same issue for me, but it could be important (or misleading) for someone trying to determine the root of the issue.
Especially, I want to thank you also for the quick AIC hack. It will serve my purposes for now, but I will wait for input from the statnet team on the meaning of ergm$likelihood before I use these values in my manuscript. And I always have the auc.pr values across the panel of networks. In fact, it may be just as good for my goal, but AIC is more familiar for most fisheries reviewers.
I am exploring both your package and STERGMs for this dynamic network, and I greatly appreciate the work that you and the statnet team have placed in your packages.
Warmest regards, Darren
Dr. Darren M. Gillis
Professor, Biological Sciences
Faculty of Science
University of Manitoba
________________________________
From: Leifeld, Philip
Sent: February 22, 2020 10:04 AM
To: statnet_help@u.washington.edu
Cc: Darren Gillis
Subject: Re: [statnet_help] Likelihood and AIC from mtergm() in the btergm package by Leifeld
Hi Darren,
While the btergm function implements a bootrapping correction to MPLE
standard errors in a TERGM (i.e., for a panel of many networks), mtergm
is just a convenience wrapper for the ergm::ergm function with MCMLE
(i.e., for few waves) based on a block-diagonal matrix with an offset
term. The original ergm object is stored inside the output, and you can
access it via mtergm.model@ergm. However, the ergm package does not seem
to define a working logLik method for models with an offset term, hence
logLik(mtergm.model@ergm) does not seem to work. Upon inspecting a
fitted model, I can see that a log likelihood value is stored in the
$loglikelihood slot. Provided that this is indeed the log likelihood,
which the statnet developers should be able to confirm, I suppose you
may be able to calculate the desired quantities as follows:
ll <- as.numeric(mtergm.model@ergm$loglikelihood)
aic <- -2 * ll + (2 * length(mtergm.model@ergm$coef))
bic <- -2 * ll + (2 * log(nobs(mtergm.model@ergm)))
That said, I am not sure how well-defined BIC is in a network context
given its reliance on the number of observations, which is perhaps
ambiguous in a network context.
If you have any follow-up questions, please feel free to email me
directly or add an entry to the issue tracker for the btergm package on
GitHub: https://github.com/leifeld/btergm/issues
Best regards,
Philip
--
Professor of Comparative Politics
Department of Government
University of Essex
http://www.philipleifeld.com
On 21/02/2020 15:13, Darren Gillis wrote:
> Hello,
>
>
> I am not sure if this is the correct place to post regarding btergm, but
> I thought I'd try as other related posts appear in the archives.
>
>
> I am currently working with a dynamic network (four annual
> cross-sections) of vessel associations in a North Sea trawl fishery. I
> am exploring both tergm::stergm() and btergm::mtergm() as alternative
> modeling methods. I am interested in using AIC to compare alternative
> models within (not between) each of these packages (in addition to
> examining MCMC chains, checking degeneracy and GOF). I can easily
> extract these values from stergm(), but in mtergm() the S4 slots for
> likelihood, aic, and bic mtergm are all NA. In addition, attempts to
> directly obtain these values with "AIC(mtergm.model) " result in:
>
>
> Error in UseMethod("logLik") : no applicable method for 'logLik' applied
> to an object of class "mtergm"
>
>
>
> Are these slots place holders for compatibility with other packages or
> have I missed something in the manual and associated publication
> (Leifeld et al. 2018)? I think the answer is pretty clear, but I
> wanted to double check before I moved on. Thank you for considering my
> query.
>
>
> Dr. Darren M. Gillis
> Professor, Biological Sciences
> Faculty of Science
> University of Manitoba
>
> ?Ignorance more frequently begets confidence than does knowledge.?
> ? Charles Darwin
>
>
> "..., reality must take precedence over public relations, for Nature
> cannot be fooled."
> ? Richard P. Feynman
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From p.krivitsky at unsw.edu.au Mon Mar 2 03:47:16 2020
From: p.krivitsky at unsw.edu.au (Pavel Krivitsky)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Likelihood and AIC from mtergm() in the btergm
package by Leifeld
In-Reply-To:
References: <9f4bf88a7c7c4165a91383c7813237b9@LNXP265MB0714.GBRP265.PROD.OUTLOOK.COM>
Message-ID: <34dc04afff2f594c7684a1ebcac54f5ced565fde.camel@unsw.edu.au>
Hi,
On Sat, 2020-02-22 at 16:04 +0000, Leifeld, Philip wrote:
> However, the ergm package does not seem
> to define a working logLik method for models with an offset term,
Are you sure? Every version I've tried handles this OK:
> logLik(ergm(flomarriage~edges+offset(edges),offset.coef=-2))
...
'log Lik.' -82.01365 (df=2)
Actually, now that I look, df should be 1 here, since there is only one
free parameter. Also, it looks like the value of the maximised
likelihood depends on the offset parameter, which, in this case, it
shouldn't:
> logLik(ergm(flomarriage~edges+offset(edges),offset.coef=-1))
...
'log Lik.' -79.6536 (df=2)
Another day, another issue (or two).
Best,
Pavel
From mheaney at umich.edu Thu Mar 5 20:58:33 2020
From: mheaney at umich.edu (Michael Heaney)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Modeling Relational Event Dynamics with R/statnet
Message-ID:
Dear Statnet Community,
Does anyone have a copy of the materials for the workshop "Modeling
Relational Event Dynamics with R/statnet" that is taught by Carter Butts?
If you could share them with me, I would appreciate it.
Also, if you happen to have any other good tutorial materials on Relational
Event Models, I would welcome them.
Thanks in advance. :)
Michael
--
Michael T. Heaney, Ph.D.
Research Fellow
School of Social and Political Sciences
University of Glasgow
Adjunct Research Professor
Institute for Research on Women and Gender
University of Michigan
email 01: Michael.Heaney@glasgow.ac.uk
email 02: mheaney@umich.edu
email 03: michaeltheaney@gmail.com
phone: +44 (0) 730 534 2286
http://www.michaeltheaney.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From buttsc at uci.edu Thu Mar 5 21:54:32 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Modeling Relational Event Dynamics with R/statnet
In-Reply-To:
References:
Message-ID:
Hi, Michael -
You can usually find fairly up-to-date workshop materials at the
workshop wiki: https://github.com/statnet/Workshops/wiki.? The
relational event workshop information is among those available.
Hope that helps!
-Carter
On 3/5/20 8:58 PM, Michael Heaney wrote:
> Dear Statnet Community,
>
> Does anyone have a copy of the materials for the workshop "Modeling
> Relational Event Dynamics with R/statnet" that is taught by Carter
> Butts?? If you could share them with me, I would appreciate it.
>
> Also, if you happen to have any other good tutorial materials on
> Relational Event Models, I would welcome them.
>
> Thanks in advance. :)
>
> Michael
>
> --
> Michael T. Heaney, Ph.D.
>
> Research Fellow
> School of Social and Political Sciences
> University of Glasgow
>
> Adjunct Research Professor
> Institute for Research on Women and Gender
> University of Michigan
>
> email 01: Michael.Heaney@glasgow.ac.uk
>
> email 02: mheaney@umich.edu
> email 03: michaeltheaney@gmail.com
>
> phone: +44 (0) 730 534 2286
> http://www.michaeltheaney.com/
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From yaxincui2023 at u.northwestern.edu Fri Mar 6 12:19:54 2020
From: yaxincui2023 at u.northwestern.edu (Yaxin Cui)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Reference distribution h(x) in valued ERGM
Message-ID:
Dear Statnet Community,
I am working on a weighted network modeling using valued-ERGM in my
research, and the determination of the choice of reference distribution
h(y).
I know there are choices such as Poisson distribution, Geometric
distribution (in ergm.count) and also Bernoulli, uniform, discrete uniform
and standard normal distribution (in ergm). Is there any mathematical form
of h(y) for each distribution? As the network structure y is the variable
in h(y) function, how does the distribution of link strength determine h(y)?
Thank you so much for your help in advance!
Regards,
Yaxin
--
Yaxin Cui, Ph.D. student
Northwestern University, Mechanical Engineering Department
Email: yaxincui2023@u.northwestern.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From malena.haenni at unisg.ch Thu Mar 12 03:21:43 2020
From: malena.haenni at unisg.ch (Haenni, Malena)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] MCMC% values
Message-ID:
Dear statnet community
I am working with valued ERGMs with count data. I am wondering what the MCMC % values mean. I never see them reported in papers, but I was told in a summer school course as a rule of thumb that the value should not be higher than 10.
Any insights on that? And what does it mean for the goodness of my models if I have a few predictors for which the values are 50 or even up to 78? This is the case for sum and one other predictor, the others are 0 ore close to 0.
What does this tell me about the goodness or the legitimacy of the models?
I would be grateful for your insights.
Best
Malena
Malena Haenni
Phd candidate
[https://owa.unisg.ch/owa/service.svc/s/GetFileAttachment?id=AAMkAGVkMjc4ZDQ2LWE2MTgtNGU2Zi1iMjM0LWMwYzM3N2RkNGM3MABGAAAAAACiVTGePE42TKTx82ovfOpiBwDZ09ADcZZYTK%2F53pderis%2BAAAAAAEJAADZ09ADcZZYTK%2F53pderis%2BAAAAAG1kAAABEgAQAHPgAW%2Fgj3xGso5x3j7BpJI%3D&X-OWA-CANARY=0fTg9RdP3kSUDpQ7yUTodZV9c_1tetQIsS4O3xbzUQzQbfvE3V3TsksfqzFbVgKFFsv07hIw8IM.]
Institut f?r Systemisches Management und Public Governance | (IMP-HSG)
Universit?t St.Gallen (HSG) | Dufourstr. 40a | CH-9000 St.Gallen
Tel. +41 79 796 50 40
malena.haenni@unisg.ch | www.imp.unisg.ch
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From p.krivitsky at unsw.edu.au Fri Mar 13 20:42:55 2020
From: p.krivitsky at unsw.edu.au (Pavel Krivitsky)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] Reference distribution h(x) in valued ERGM
In-Reply-To:
References:
Message-ID: <9b5553ef69b1693b5077d0a37429a69191d814f3.camel@unsw.edu.au>
Dear Yaxin,
It should be in the help for the references:
help('ergm-references')
Best,
Pavel
On Fri, 2020-03-06 at 14:19 -0600, Yaxin Cui wrote:
Dear Statnet Community,
I am working on a weighted network modeling using valued-ERGM in my research, and the determination of the choice of reference distribution h(y).
I know there are choices such as Poisson distribution, Geometric distribution (in ergm.count) and also Bernoulli, uniform, discrete uniform and standard normal distribution (in ergm). Is there any mathematical form of h(y) for each distribution? As the network structure y is the variable in h(y) function, how does the distribution of link strength determine h(y)?
Thank you so much for your help in advance!
Regards,
Yaxin
--
Yaxin Cui, Ph.D. student
Northwestern University, Mechanical Engineering Department
Email: yaxincui2023@u.northwestern.edu
_______________________________________________
statnet_help mailing list
statnet_help@u.washington.edu
http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From p.krivitsky at unsw.edu.au Fri Mar 13 20:48:31 2020
From: p.krivitsky at unsw.edu.au (Pavel Krivitsky)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] MCMC% values
In-Reply-To:
References:
Message-ID: <01696a3554422dc8c5065278587c2c03b2185ba4.camel@unsw.edu.au>
Dear Malena,
MCMC % is the percent of the standard error in the table that is due to using MCMC to estimate the likelihood (loosely speaking). If the model converages but the MCMC % is that high, it means that the model is not too problematic, but the MCMC.interval= control.ergm() parameter is too small to get an accurate idea of what the MLE is.
You can play around with it, or you can try the MCMC.effectiveSize= mechanism, which will try to find it automatically---though it's relatively untested.
I hope this helps,
Pavel
On Thu, 2020-03-12 at 10:21 +0000, Haenni, Malena wrote:
Dear statnet community
I am working with valued ERGMs with count data. I am wondering what the MCMC % values mean. I never see them reported in papers, but I was told in a summer school course as a rule of thumb that the value should not be higher than 10.
Any insights on that? And what does it mean for the goodness of my models if I have a few predictors for which the values are 50 or even up to 78? This is the case for sum and one other predictor, the others are 0 ore close to 0.
What does this tell me about the goodness or the legitimacy of the models?
I would be grateful for your insights.
Best
Malena
Malena Haenni
Phd candidate
[https://owa.unisg.ch/owa/service.svc/s/GetFileAttachment?id=AAMkAGVkMjc4ZDQ2LWE2MTgtNGU2Zi1iMjM0LWMwYzM3N2RkNGM3MABGAAAAAACiVTGePE42TKTx82ovfOpiBwDZ09ADcZZYTK%2F53pderis%2BAAAAAAEJAADZ09ADcZZYTK%2F53pderis%2BAAAAAG1kAAABEgAQAHPgAW%2Fgj3xGso5x3j7BpJI%3D&X-OWA-CANARY=0fTg9RdP3kSUDpQ7yUTodZV9c_1tetQIsS4O3xbzUQzQbfvE3V3TsksfqzFbVgKFFsv07hIw8IM.]
Institut f?r Systemisches Management und Public Governance | (IMP-HSG)
Universit?t St.Gallen (HSG) | Dufourstr. 40a | CH-9000 St.Gallen
Tel. +41 79 796 50 40
malena.haenni@unisg.ch | www.imp.unisg.ch
_______________________________________________
statnet_help mailing list
statnet_help@u.washington.edu
http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From morrism at uw.edu Sat Mar 21 09:16:37 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] An error on gplot.target() function (fwd)
Message-ID:
---------- Forwarded message ----------
Date: Sat, 21 Mar 2020 12:05:20 +0000
From: hakan hekim
To: "morrism@uw.edu"
Subject: An error on gplot.target() function
Hi,
I am working on a SNA project and using statnet. I am having an error, I couldn't find any contact information on statnet
website. I found your e-mail at statnet package document on CRAN.
Following code;
sna::gplot.target(g, closeness(g), main="Degree", ?circ.lab = FALSE,
? ? ? ? ? ? ? ? ? circ.col="skyblue", usearrows = FALSE,
? ? ? ? ? ? ? ? ? circ.lab.col = "white",
? ? ? ? ? ? ? ? ? vertex.col=c("blue", rep("red", 32), "yellow"),
? ? ? ? ? ? ? ? ? edge.col="darkgray", displaylabels = TRUE,
? ? ? ? ? ? ? ? ? label = netfriends_myk %v% "alias", label.cex=0.5,)
develops this warning;
In doTryCatch(return(expr), name, parentenv, handler) :? "circ.lab" is not a graphical parameter
and in some cases does not draw the plot.
Can you help me on this issue or point to someone who can help?
Thanks.
Hakan
From morrism at uw.edu Sat Mar 21 10:31:36 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:14 2021
Subject: [statnet_help] An error on gplot.target() function (fwd)
Message-ID:
Hi,
I am working on a SNA project and using statnet. I am having an error, I
couldn't find any contact information on statnet
website. I found your e-mail at statnet package document on CRAN.
Following code;
sna::gplot.target(g, closeness(g), main="Degree", ?circ.lab = FALSE,
? ? ? ? ? ? ? ? ? circ.col="skyblue", usearrows = FALSE,
? ? ? ? ? ? ? ? ? circ.lab.col = "white",
? ? ? ? ? ? ? ? ? vertex.col=c("blue", rep("red", 32), "yellow"),
? ? ? ? ? ? ? ? ? edge.col="darkgray", displaylabels = TRUE,
? ? ? ? ? ? ? ? ? label = netfriends_myk %v% "alias", label.cex=0.5,)
develops this warning;
In doTryCatch(return(expr), name, parentenv, handler) :? "circ.lab" is not a
graphical parameter
and in some cases does not draw the plot.
Can you help me on this issue or point to someone who can help?
Thanks.
Hakan
From buttsc at uci.edu Sat Mar 21 11:53:18 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] An error on gplot.target() function (fwd)
In-Reply-To:
References:
Message-ID: <418ef0a8-040b-1bd0-2f27-ee8b500d0ff7@uci.edu>
Hi, Hakan (by way of Martina):
A fix will be in the next sna release, but in the meantime I've attached
a modified version of gplot.target (suitable for source()ing) that will
fix the spurious warnings regarding circ.lab.? I'm not able to reproduce
the failure to draw the plot, however.? If there's a reproducible
example of that, I'll take a look.
Hope that helps,
-Carter
On 3/21/20 10:31 AM, martina morris wrote:
>
> Hi,
>
> I am working on a SNA project and using statnet. I am having an error,
> I couldn't find any contact information on statnet
> website. I found your e-mail at statnet package document on CRAN.
>
> Following code;
> sna::gplot.target(g, closeness(g), main="Degree", ?circ.lab = FALSE,
> ? ? ? ? ? ? ? ? ? circ.col="skyblue", usearrows = FALSE,
> ? ? ? ? ? ? ? ? ? circ.lab.col = "white",
> ? ? ? ? ? ? ? ? ? vertex.col=c("blue", rep("red", 32), "yellow"),
> ? ? ? ? ? ? ? ? ? edge.col="darkgray", displaylabels = TRUE,
> ? ? ? ? ? ? ? ? ? label = netfriends_myk %v% "alias", label.cex=0.5,)
>
> develops this warning;
> In doTryCatch(return(expr), name, parentenv, handler) : "circ.lab" is
> not a graphical parameter
>
> and in some cases does not draw the plot.
>
> Can you help me on this issue or point to someone who can help?
>
> Thanks.
>
> Hakan
>
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
#Modified version of gplot.target that fixes a spurious warning; to appear in
#sna version 2.6
#
#CTB, 3/21/20
#
gplot.target<-
function (dat, x, circ.rad = (1:10)/10, circ.col = "blue", circ.lwd = 1,
circ.lty = 3, circ.lab = TRUE, circ.lab.cex = 0.75, circ.lab.theta = pi,
circ.lab.col = 1, circ.lab.digits = 1, circ.lab.offset = 0.025,
periph.outside = FALSE, periph.outside.offset = 1.2, ...)
{
offset <- min(0.5, sum(x == max(x))/(length(x) - 1))
xrange <- diff(range(x))
xmin <- min(x)
x <- 1 - (x - xmin)/(xrange + offset)
circ.val <- (1 - circ.rad) * (xrange + offset) + xmin
cl <- match.call()
if (is.null(cl$layout.par))
cl$layout.par <- list(radii = x)
else cl$layout.par$radii <- x
cl$layout.par$periph.outside <- periph.outside
cl$layout.par$periph.outside.offset <- periph.outside.offset
cl$x <- NULL
cl$circ.rad <- NULL
cl$circ.col <- NULL
cl$circ.lwd <- NULL
cl$circ.lty <- NULL
cl$circ.lab.theta <- NULL
cl$circ.lab.col <- NULL
cl$circ.lab.cex <- NULL
cl$circ.lab.digits <- NULL
cl$circ.lab.offset <- NULL
cl$periph.outside <- NULL
cl$periph.outside.offset <- NULL
cl$circ.lab <- NULL
cl$mode <- "target"
cl$xlim = c(-periph.outside.offset, periph.outside.offset)
cl$ylim = c(-periph.outside.offset, periph.outside.offset)
cl[[1]] <- match.fun("gplot")
coord <- eval(cl)
if (length(circ.col) < length(x))
circ.col <- rep(circ.col, length = length(x))
if (length(circ.lwd) < length(x))
circ.lwd <- rep(circ.lwd, length = length(x))
if (length(circ.lty) < length(x))
circ.lty <- rep(circ.lty, length = length(x))
for (i in 1:length(circ.rad)) segments(circ.rad[i] * sin(2 *
pi/100 * (0:99)), circ.rad[i] * cos(2 * pi/100 * (0:99)),
circ.rad[i] * sin(2 * pi/100 * (1:100)), circ.rad[i] *
cos(2 * pi/100 * (1:100)), col = circ.col[i], lwd = circ.lwd[i],
lty = circ.lty[i])
if (circ.lab)
text((circ.rad + circ.lab.offset) * cos(circ.lab.theta),
(circ.rad + circ.lab.offset) * sin(circ.lab.theta),
round(circ.val, digits = circ.lab.digits), cex = circ.lab.cex,
col = circ.lab.col)
invisible(coord)
}
From mheaney at umich.edu Wed Mar 25 17:58:02 2020
From: mheaney at umich.edu (Michael Heaney)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Online course on Social Network Analysis (June 8-12,
2020)
Message-ID:
STATISTICAL ANALYSIS OF SOCIAL NETWORK DATA (Please share!) -- I am
offering an intensive, one-week, online, graduate-level course on social
network analysis through the University of St. Gallen, Switzerland, from
June 8-12, 2020. We will cover the fundamentals of both descriptive (e.g.,
centrality, visualization) and inferential network analysis (e.g., ERGM,
TERGM). This will be a small course capped at 20 students, so students will
be guaranteed individualized attention through class sessions and abundant
online office hours. Students who successfully complete the course will
receive 4 ECTS (European Credit Transfer and Accumulation System). Direct
questions to michaeltheaney@gmail.com. More info here:
https://www.gserm.ch/stgallen/course/?course_code=10,828
Michael
--
Michael T. Heaney, Ph.D.
Adjunct Research Professor
Institute for Research on Women and Gender
University of Michigan
1136 Lane Hall
204 S. State Street
Ann Arbor, MI 48109-1290 USA
Phone: +1.734.764.9537
E-mail: mheaney@umich.edu
http://www.michaeltheaney.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From goodreau at uw.edu Thu Apr 9 15:39:32 2020
From: goodreau at uw.edu (Steven Goodreau)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Network Modeling for Epidemics -- Summer short
course at the University of Washington -- 17-21 August, 2020 - UPDATES
In-Reply-To: <17c3ec23-f776-bdd7-ba4d-efdb41c6f856@uw.edu>
References:
<17c3ec23-f776-bdd7-ba4d-efdb41c6f856@uw.edu>
Message-ID: <56de3909-4da9-449d-6f4f-229752479a03@uw.edu>
Hello all -
A reminder about our summer course, and a COVID-19 update.
Best,
Steve Goodreau, Martina Morris, Sam Jenness
############################################################################
April 2020 COVID-19 Update
Due to the travel restrictions associated with the COVID-19 global
pandemic, we have updated our plans for NME 2020 as follows:
In addition to the standard in-person workshop format, we will be
offering a concurrent virtual, web-based option for participation in NME
2020. This web-based option will be hosted live via Zoom (or related
technology), concurrent with the in-person workshop. Virtual students
will access our lectures live (during our regular class hours) and/or in
recorded form, complete and submit the interactive lab activities online
for instructor comment and evaluation, and discuss the material with the
instructors via video-based and text-based chats. Although the class
interactions will be more limited with the virtual option, the
instructors will
make every effort to deliver and discuss the material collectively to
engage both in-person and virtual students. If you are interested in the
virtual option, please indicate this on your application.
We will continue to monitor whether we can hold the in-person NME 2020
due to ongoing travel restrictions over the next 2 months. By June 1, we
will make the final determination about whether the in-person NME 2020
will be held. If an in-person NME is not possible, we will conduct NME
2020 fully in a virtual format. We will contact everyone registered for
NME 2020 by email on June 1 with our decision.
On 2/17/2020 7:15 PM, Steven Goodreau wrote:
> ############################################################################
>
>
> Network Modeling for Epidemics
> Summer short course at the University of Washington
> 17-21 August, 2020
>
> Network Modeling for Epidemics (NME) is a 5-day short course at the
> University of Washington that provides an introduction to stochastic
> network models for infectious disease transmission dynamics, with a
> focus on empirically based modeling of HIV transmission. It is a
> ''hands-on'' course, using the EpiModel software package in R
> (www.epimodel.org). EpiModel provides a unified framework for
> statistically based modeling of dynamic networks from empirical data,
> and simulation of epidemic dynamics on these networks. It has a
> flexible open-source platform for learning and building several types
> of epidemic models: deterministic compartmental, stochastic
> individual-based, and stochastic network models. Resources include
> simple models that run in a browser window, built-in generic models
> that provide basic control over population contact patterns, pathogen
> properties and demographics, and templates for user-programmed modules
> that allow EpiModel to be extended to the full range of pathogens,
> hosts, and disease dynamics for advanced research. This course will
> touch on the deterministic and individual-based models, but its
> primary focus is on the theory, methods and application of network
> models.
>
> The course uses a mix of lectures, tutorials, and labs with students
> working in small groups. On the final day, students work to develop an
> EpiModel prototype model (either individually or in groups based on
> shared research interests), with input from the instructors, including
> the lead EpiModel software developer, Dr. Samuel Jenness.
>
> Returning students: We encourage previous attendees with active
> modeling projects to apply to return for a refresher course. The
> EpiModel package has been significantly enhanced over the last few
> years. Returning students with active projects will have the
> opportunity to work with course instructors to address key challenges
> in the design of their network model code.
>
> *Dates and location:*
>
> The course will be taught from Monday, August 17 to Friday, August 21
> on the University of Washington campus in Seattle.
>
> *Costs:*
>
> Course fee is $800. Travel and accommodation costs are the
> responsibility of the participant, although discounted hotel rates are
> available. We offer a limited number of fee waivers for pre-doctoral
> students or for attendees from low income countries. These cover
> waiver of the registration fee only; travel and accommodation are
> still the responsibility of the fee waiver recipient.
>
> *Application dates and decision dates:*
>
> *??? Apr 19: Fee-waiver application deadline. Decisions will be made
> by Apr 30, and response required by May 15.
>
> *??? May 17:? General application deadline. Decisions will be made by
> May 31 and a response required by June 15.
>
> *??? A waitlist will be established along with rolling admission
> through June 30 if space allows.
>
> *Application:*
>
> Apply online at https://catalyst.uw.edu/webq/survey/morrism/385149
>
> Course website and more information: http://statnet.github.io/nme
>
> Please free to share widely!
>
> Note: We will also be teaching this course in Antwerp, Belgium, 7-11
> September.? More information can be found here:
> https://www.uantwerpen.be/en/summer-schools/modelling-infectious-diseases/
>
> Yours,
> Martina Morris, Steve Goodreau and Samuel Jenness
>
--
*****************************************************************
Steven M. Goodreau / Professor / Dept. of Anthropology
Physical address: Denny Hall M236
Mailing address: Campus Box 353100 / 4216 Memorial Way NE
Univ. of Washington / Seattle WA 98195
1-206-685-3870 (phone) /1-206-543-3285 (fax)
http://faculty.washington.edu/goodreau
*****************************************************************
From mheaney at umich.edu Tue Apr 14 15:31:04 2020
From: mheaney at umich.edu (Michael Heaney)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Troubleshooting for Relational Event Models
In-Reply-To:
References:
Message-ID:
Dear Statnet Community,
I would most appreciate any assistance in troubleshooting for a Relational
Event Model.
I am mimicking Section 3 of "Modeling Relational Event Dyanmics with
statnet", which is available here: https://github.com/statnet/Workshops/wiki
The code works fine when I use the data provided by the workshop. I only
get the error when using my own data.
Everything works in the code until the semifinal line, after which I get
the following error:
Error in check.objfun.output(out, minimize, d) :
objfun returned value that is NA or NaN
I have posted my adapted code below. The problem is evident after the line
that starts with gamma.fit2<-rem( . . .
Here is the code:
# Open Libraries
library(statnet)
library(relevent)
library(informR)
# Read Data
data <- read.csv("Hashtags_More_Than_Once_and_Attributes_2020-04-12.csv",
header=TRUE,stringsAsFactors=FALSE)
rawevents<-cbind(data$handle, data$hashtag)
evls<-gen.evl(rawevents)
names(evls) # See the structure of the evls object
length(evls$eventlist) # N
evls$eventlist[[1]] # The first event history in the eventlist
evls$event.key # The event.key maps the tokens to the event type names.
alpha.ints <- gen.intercepts(evls)
# Examine the structure of the alpha.ints statslist
length(alpha.ints) # N
dim(alpha.ints[[1]][[1]])
alpha.ints[[1]][[1]][1,,]
alpha.fit<-rem(eventlist=evls$eventlist,statslist=alpha.ints,
estimator="BPM",prior.param=list(mu=0,sigma=100,nu=4))
summary(alpha.fit)
pois.mle<-log(prop.table(table(data$handle))[-c(7,10)]
/prop.table(table(data$handle))[10])
round(cbind(BPM=alpha.fit$coef[order(names(alpha.fit$coef))],pois.mle),4)
a1<-paste(evls$event.key[-9,1],evls$event.key[-9,1],sep="") # All inertial
terms
a1
set.seed(12345)
inds<-sample(1:length(evls$eventlist),500)
evls$eventlist<-evls$eventlist[inds]
alpha.ints<-alpha.ints[inds]
beta.sforms<-gen.sformlist(evls,a1)
evls$eventlist$trumprussia
beta.sforms$"trumprussia"[1:4,,c("aa","bb")]
beta.ints<-slbind(beta.sforms,alpha.ints,type=1,new.names=TRUE,event.key=evls$event.key)
beta.fit<-rem(evls$eventlist,beta.ints,estimator="BPM",
prior.param=list(mu=0,sigma=100,nu=4))
round(cbind(BPM=beta.fit$coef[13:25],Z=beta.fit$coef[13:25]/beta.fit$sd[13:25]),4)
beta.sansints<-sfl2statslist(beta.sforms)
a2<-c("a(c|h)b","ad+b")
gamma.sforms<-gen.sformlist(evls,a2)
gamma.ints<-slbind(gamma.sforms,beta.ints,new.names=TRUE,event.key=evls$event.key)
gamma.fit<-rem(evls$eventlist,gamma.ints,estimator="BPM",
prior.param=list(mu=0,sigma=100,nu=4))
summary(gamma.fit)
moveon.sfs<-paste("a",letters[2:14],"a",sep="")
moveon.sfs
moveon.sforms<-glb.sformlist(evls,sforms=list(moveon.sfs),new.names="InterMoveOn")
moveon.ints<-slbind(moveon.sforms, gamma.ints)
moveon.fit<-rem(evls$eventlist,moveon.ints,estimator="BPM",
prior.param=list(mu=0,sigma=100,nu=4))
round(cbind(BPM=moveon.fit$coef,Z=moveon.fit$coef/moveon.fit$sd)[26:28,],4)
delta.ints<-sldrop(moveon.ints,varname=c("InterMoveOn","publicipublici"))
delta.fit<-rem(evls$eventlist,delta.ints) # fit it.
names(delta.fit$coef)
c(delta.fit$BIC,moveon.fit$BIC)
sl.ind<-87:88
New.evs<-unlist(glapply(data$New,data$hashtag,unique,regroup=FALSE))
New.evs<-ifelse(New.evs==1,1,0)[inds]
gamma.ints2<-slbind.cond(New.evs,gamma.ints,sl.ind=sl.ind,var.suffix="NEW")
sl.ind<-87:90
gamma.fit2<-rem(evls$eventlist,gamma.ints2,estimator="BPM", prior.param=
list(mu=0,sigma=100,nu=4))
round(cbind(BPM=gamma.fit2$coef[sl.ind],Z=gamma.fit2$coef[sl.ind]/gamma.fit2$sd[sl.ind]),4)
--
Michael T. Heaney, Ph.D.
Adjunct Research Professor
Institute for Research on Women and Gender
University of Michigan
1136 Lane Hall
204 S. State Street
Ann Arbor, MI 48109-1290 USA
Phone: +1.734.764.9537
E-mail: mheaney@umich.edu
http://www.michaeltheaney.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From pian.wang at szu.edu.cn Mon Apr 20 05:47:28 2020
From: pian.wang at szu.edu.cn (=?UTF-8?B?5rGq57+p57+p?=)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] issues with dyadcov # Error in solve.default(H,
tol = 1e-20) : Lapack routine dgesv: system is exactly singular:
U[11, 11] = 0
Message-ID: <479435b.3ea7d.17197a1425d.Coremail.pian.wang@szu.edu.cn>
Hello statnet community,
I have a question regarding to ?dyadcov?. I?m trying to use the term ?dyadcov? to test the influence of a follower-followee network of a group of people on the formation of their communication network. But several error messages popping up during this process.
The follower-followee network (ff) and communication network (g) are both directed network. The communication network are also weighted network. I intend to see the whether the presence of a tie in follower-followee network in a dyad would impact the presence of a tie in their communication network. So I guest the dyadcov is the proper term to model this, right?
Everything looks fine when I put only node attribute terms (e.g. nodeicov (?fan?)) and homophilic attribute (e.g. nodematch (?location?)) and dyadcov term dyadcov(ff) into the model.
fit0.1<-ergm(g~edges
+nodeicov("follow")+nodeicov("weibo")+nodeicov("fans")
+nodematch("location")+nodematch("group")+nodematch("level")
+dyadcov(ff),
control=control.ergm(MCMC.burnin=10000,MCMC.samplesize = 5000,seed=1,MCMC.interval=10000,parallel = 4,parallel.type = "PSOCK"),
response = NULL, verbose = FALSE)
#This model was fit using MCMC. To examine model diagnostics and check for degeneracy, use the mcmc.diagnostics() function.
However, when I add an endogenous term ?mutual? into the model
fit0.2<-ergm(g~edges
+nodeicov("follow")+nodeicov("weibo")+nodeicov("fans")
+nodematch("location")+nodematch("group")+nodematch("level")
+dyadcov(ff)
+mutual,
control=control.ergm(MCMC.burnin=10000,MCMC.samplesize = 5000,seed=1,MCMC.interval=10000,parallel = 4,parallel.type = "PSOCK"),
response = NULL, verbose = FALSE)
#The model was converged once after Iteration 1.
But it returns error at Iteration 2.
#Iteration 2 of at most 20:
#Optimizing with step length 1.
#Error in solve.default(H, tol = 1e-20) :
Lapack routine dgesv: system is exactly singular: U[11,11] = 0
What does this error mean? How can I solve this problem?
Actually I realize that it maybe the combination of two terms ?dyadcov? and ?mutual? that cause the problem, because when I replace mutual with other terms, it works fine.
With the combination of dyadcov and some other structural terms (e.g. dgwdsp, gwidegree) the model can be converged, but still I got the following warning messages:
# in term ?dyadcov? in package ?ergm?: asymmetric covariate in dyadcov using upper triangle only. (and the same appears repeatedly in one run)
I can see all the model coefficients are all reported in some cases, including lower triangle, but missing of lower coefficients in other cases, i.e. only upper triangle coefficients along with other coefficients are reported. Wondering why it happens?
I hope my messages won?t be too long to read .
Thank you in advance for any advice on what to look out for!
Best regards,
Dr. Pianpian Wang
Assistant Professor
School of media and communication
Shenzhen University
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From marc.sarazin at uclouvain.be Thu Apr 23 23:49:54 2020
From: marc.sarazin at uclouvain.be (Marc Sarazin)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] DGWESP/DGWDSP terms with node attributes
Message-ID:
Dear statnetters,
I hope you are all well and staying safe.
In a nutshell: has anyone come up with versions of directed GWESP/GWDSP terms that only take into account nodes with the same values for a categorical attribute (e.g. ? la ttriple(attr = "attribute", diff = TRUE))?
The longer version: I am attempting to model the presence of different kinds of triangles among nodes that share the same categorical attribute in my network, given the general presence of triangular structures. I have tried fitting models with triangle or ttriple terms specifically for the attribute (e.g. triangle(attr = "attribute", diff = TRUE), alongside a general directed GWESP term (path closure configuration, with fixed decay). Unfortunately (and unsurprisingly) the models don't come close to converging. The same model without the triangle/ttriple terms fits perfectly well.
My suspicion is that this is due to the non-convergence issues encountered with Markov dependence terms. Therefore, I'd like to use a GWESP term that counts only edgewise shared partners with the same nodal attribute instead, but this term isn't currently available. I was wondering then if anyone had developed such an attribute-based GWESP term? Eventually I'll want to look at closure processes as well-so has anyone equally developed an equivalent GWDSP term? Or are there known problems that make developing such terms difficult? Alternatively, has anyone come up with workarounds to using such terms (that don't involve splitting the network according to the attribute in question-which is of course not ideal)? For info, my attribute divides the nodes of my network evenly into 5 groups (it is based on quintiles of a continuous quantitative attribute)
Very best wishes,
Marc
--
Dr Marc Sarazin
Postdoctoral Research Fellow
GIRSEF, UCLouvain (University of Louvain/Universit? catholique de Louvain)
+32 (0)10479446 | Project Website
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From goodreau at uw.edu Mon May 11 19:37:48 2020
From: goodreau at uw.edu (Steven Goodreau)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] DGWESP/DGWDSP terms with node attributes
In-Reply-To:
References:
Message-ID: <63ba345d-bbb0-3a72-f796-8500ae197f05@uw.edu>
Hi Marc -
I don't know of anyone who has coded those up (that I can think of -
another member of the statnet team, correct me if I'm wrong).? They seem
like they would be quite useful additions.
Just FYI, the latest versions of our tutorials for coding your own terms
are at:
https://github.com/statnet/Workshops/wiki
down at the bottom.
Best,
Steve
On 4/23/2020 11:49 PM, Marc Sarazin wrote:
>
> Dear statnetters,
>
> I hope you are all well and staying safe.
>
> In a nutshell: has anyone come up with versions of directed
> GWESP/GWDSP terms that only take into account nodes with the same
> values for a categorical attribute (e.g. ? la ttriple(attr =
> ?attribute?, diff = TRUE))?
>
> The longer version: I am attempting to model the presence of different
> kinds of triangles among nodes that share the same categorical
> attribute in my network, given the general presence of triangular
> structures. I have tried fitting models with triangle or ttriple terms
> specifically for the attribute (e.g. triangle(attr = ?attribute?, diff
> = TRUE), alongside a general directed GWESP term (path closure
> configuration, with fixed decay). Unfortunately (and unsurprisingly)
> the models don?t come close to converging. The same model without the
> triangle/ttriple terms fits perfectly well.
>
> My suspicion is that this is due to the non-convergence issues
> encountered with Markov dependence terms. Therefore, I?d like to use a
> GWESP term that counts only edgewise shared partners with the same
> nodal attribute instead, but this term isn?t currently available. I
> was wondering then if anyone had developed such an attribute-based
> GWESP term? Eventually I?ll want to look at closure processes as
> well?so has anyone equally developed an equivalent GWDSP term? Or are
> there known problems that make developing such terms difficult?
> Alternatively, has anyone come up with workarounds to using such terms
> (that don?t involve splitting the network according to the attribute
> in question?which is of course not ideal)? For info, my attribute
> divides the nodes of my network evenly into 5 groups (it is based on
> quintiles of a continuous quantitative attribute)
>
> Very best wishes,
>
> Marc
>
> --
>
> Dr Marc Sarazin
>
> Postdoctoral Research Fellow
>
> GIRSEF, UCLouvain (University of Louvain/Universit? catholique de Louvain)
>
> +32 (0)10479446 | Project Website
>
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
--
*****************************************************************
Steven M. Goodreau / Professor / Dept. of Anthropology
Physical address: Denny Hall M236
Mailing address: Campus Box 353100 / 4216 Memorial Way NE
Univ. of Washington / Seattle WA 98195
1-206-685-3870 (phone) /1-206-543-3285 (fax)
http://faculty.washington.edu/goodreau
*****************************************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From kraft.tom at gmail.com Tue May 12 08:21:04 2020
From: kraft.tom at gmail.com (Tom Kraft)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
Message-ID:
Dear statnet,
Thank you for all your wonderful work. I am looking for advice on methods
applicable to network data that doesn't fit neatly into the types of data
that regularly appear in statnet tutorials. I have a list of sequential
discrete time networks involving the same set of actors. At each timepoint,
however, I only observe some percent of actors in the whole network (~30%).
For all actors observed I have complete knowledge of their ties, including
the identity of alters. I would like construct a model of these empirical
data with relatively basic terms that can be used to simulate full temporal
networks with similar network structures.
Given the temporal nature of the networks, it seems that a standard tergm
could appropriately be used to model the formation and dissolution of ties.
Yet given the nature of the sampling, for any given network I'm inclined to
think that the data are essentially egocentric with known alter info
because if I have information on a node I can be sure that I know all the
connections of that individual. Thus, I am wondering if it is possible to
conduct a temporal version of the ergm.ego model. From the materials I have
found online, it seems this should be possible:
"The principles of egocentric inference can be extended to temporal ERGMs
(TERGMs). While we will not cover that in this workshop, an example can be
found in another paper that sought to evaluate the network hypothsis for
racial disparities in HIV in the US (Morris et al. 2009)." -
http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
However, I am unable to find worked examples to follow up on that
reference/approach and it is not clear to me how to implement this.
Additionally, it seems like ergm.ego is not designed to incorporate
information on identifiable alters -- in which case the methods developed
in Kosikinen and Robins (2010) perhaps would be more appropriate so that
this useful information is not ignored.
Alternatively, I could imagine that this analysis is best conceived of as a
tergm with missing data at each time step. In this case info on alter ids
could be fully utilized I think.
I would be very grateful if anyone could comment on whether one of these
approaches seems feasible, or if there are other options I might consider.
Any materials or references to vignettes/tutorials/papers on the topic
would also be appreciated. Thank you in advance! Best,
Tom
Thomas Kraft
Postdoctoral Scholar
Department of Anthropology
University of California, Santa Barbara
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From kraft.tom at gmail.com Tue May 12 12:19:11 2020
From: kraft.tom at gmail.com (Tom Kraft)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
In-Reply-To:
References:
Message-ID:
Hi Zack,
Thank you for this helpful reference. I had come across this paper in my
literature search but should have given it more attention.
Is there a place where the source code associated with your paper is
available? It remains a bit unclear to me how several of the strategies you
used (Shat approximation) might be implemented in R.
I suppose an alternative is to use multiple imputation as discussed here:
Wang, C., Butts, C. T., Hipp, J. R., Jose, R., & Lakon, C. M. (2016).
Multiple imputation for missing edge data: a predictive evaluation method
with application to add health. *Social networks*, *45*, 89-98.
Thank you!
Tom
On Tue, May 12, 2020 at 1:26 PM Zack Almquist wrote:
> Hi Thomas,
>
> My paper on missing data for TERGM style analysis might be helpful; take a
> look at it and see if it is addressing your problem of interest:
>
> Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis with
> missing data: theory and methods." *Statistica Sinica* 28.3 (2018):
> 1245-1264.
>
> Best,
>
> Zack
> ---
> Zack W. Almquist
> Assistant Professor
> Department of Sociology
> Senior Data Scientist Fellow, eScience Institute
> University of Washington
>
>
> On Tue, May 12, 2020 at 8:25 AM Tom Kraft wrote:
>
>> Dear statnet,
>>
>> Thank you for all your wonderful work. I am looking for advice on methods
>> applicable to network data that doesn't fit neatly into the types of data
>> that regularly appear in statnet tutorials. I have a list of sequential
>> discrete time networks involving the same set of actors. At each timepoint,
>> however, I only observe some percent of actors in the whole network (~30%).
>> For all actors observed I have complete knowledge of their ties, including
>> the identity of alters. I would like construct a model of these empirical
>> data with relatively basic terms that can be used to simulate full temporal
>> networks with similar network structures.
>>
>> Given the temporal nature of the networks, it seems that a standard tergm
>> could appropriately be used to model the formation and dissolution of ties.
>> Yet given the nature of the sampling, for any given network I'm inclined to
>> think that the data are essentially egocentric with known alter info
>> because if I have information on a node I can be sure that I know all the
>> connections of that individual. Thus, I am wondering if it is possible to
>> conduct a temporal version of the ergm.ego model. From the materials I have
>> found online, it seems this should be possible:
>>
>> "The principles of egocentric inference can be extended to temporal
>> ERGMs (TERGMs). While we will not cover that in this workshop, an example
>> can be found in another paper that sought to evaluate the network hypothsis
>> for racial disparities in HIV in the US (Morris et al. 2009)." -
>> http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
>>
>> However, I am unable to find worked examples to follow up on that
>> reference/approach and it is not clear to me how to implement this.
>> Additionally, it seems like ergm.ego is not designed to incorporate
>> information on identifiable alters -- in which case the methods developed
>> in Kosikinen and Robins (2010) perhaps would be more appropriate so that
>> this useful information is not ignored.
>>
>> Alternatively, I could imagine that this analysis is best conceived of as
>> a tergm with missing data at each time step. In this case info on alter ids
>> could be fully utilized I think.
>>
>> I would be very grateful if anyone could comment on whether one of these
>> approaches seems feasible, or if there are other options I might consider.
>> Any materials or references to vignettes/tutorials/papers on the topic
>> would also be appreciated. Thank you in advance! Best,
>>
>> Tom
>>
>> Thomas Kraft
>> Postdoctoral Scholar
>> Department of Anthropology
>> University of California, Santa Barbara
>> _______________________________________________
>> statnet_help mailing list
>> statnet_help@u.washington.edu
>> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From c.e.g.steglich at rug.nl Tue May 12 23:29:19 2020
From: c.e.g.steglich at rug.nl (Christian Steglich)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
In-Reply-To:
References:
Message-ID: <20b71340-00d9-bf88-a5b7-a59c99085cd7@rug.nl>
Dear Tom and Zack,
I'd like to also point to highly related work by Robert Krause on
missing data in ergm and stochastic actor-oriented network models (saom):
*
Krause, R. W., Huisman, M., & Snijders, T. A. (2018). Multiple
imputation for longitudinal network data. /Italian Journal of
Applied Statistics/, /30/(30), 33-58.
*
Krause, R. W., Huisman, M., Steglich, C., & Sniiders, T. A. (2018,
August). Missing network data: A comparison of different imputation
methods. In /2018 IEEE/ACM International Conference on Advances in
Social Networks Analysis and Mining (ASONAM)/ (pp. 159-163). IEEE.
*
Krause, R. W., Huisman, M., Steglich, C., & Snijders, T. (2020).
Missing data in cross-sectional networks?An extensive comparison of
missing data treatment methods. /Social Networks/, /62/, 99-112.
For the choice between discrete-time approaches (like tergm) and
continuous-time models (like saom), it can be helpful to consider
* whether time intervals are very small compared to the network
evolution speed (then both approaches may make sense but will not
differ much from each other, nor from normal, independence-assuming,
logistic regression*), and
* whether observation moments are clearly round-based (e.g., derived
from yearly publications) or clearly snapshots of a continuous-time
process (e.g., repeatedly measuring friendship in an organisation) -
see reference** below.
Like in the STERGM- approach, also in the saom-framework, the difference
between tie creation and tie dissolution is modelled by separate
functions (keyword: endowment, creation, evaluation functions - see
RSiena Manual***, p.13f).
All the best, Christian
* Lerner, J., Indlekofer, N., Nick, B., & Brandes, U. (2013).
Conditional independence in dynamic networks. /Journal of Mathematical
Psychology/, /57/(6), 275-283.
** Block, P., Koskinen, J., Hollway, J., Steglich, C., & Stadtfeld, C.
(2018). Change we can believe in: Comparing longitudinal network models
on consistency, interpretability and predictive power. /Social
Networks/, /52/, 180-191.
*** http://www.stats.ox.ac.uk/~snijders/siena/RSiena_Manual.pdf
On 12/05/2020 21:19, Tom Kraft wrote:
> Hi Zack,
>
> Thank you for this helpful reference. I had come across this paper in
> my literature search but should have given it more attention.
>
> Is there a place where the source code associated with your paper is
> available? It remains a bit unclear to me how several of the
> strategies you used (Shat?approximation) might be implemented in R.
>
> I suppose an alternative is to use multiple imputation as discussed here:
> Wang, C., Butts, C. T., Hipp, J. R., Jose, R., & Lakon, C. M. (2016).
> Multiple imputation for missing edge data: a predictive evaluation
> method with application to add health. /Social networks/, /45/, 89-98.
>
> Thank you!
> Tom
>
> On Tue, May 12, 2020 at 1:26 PM Zack Almquist > wrote:
>
> Hi Thomas,
>
> My paper on missing data for TERGM style analysis might be
> helpful; take a look at it and see if it is addressing your
> problem of interest:
>
> Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis
> with missing data: theory and methods." /Statistica Sinica/?28.3
> (2018): 1245-1264.
>
> Best,
>
> Zack
> ---
> Zack W. Almquist
> Assistant Professor
> Department of Sociology
> Senior Data Scientist Fellow, eScience Institute
> University of Washington
>
>
> On Tue, May 12, 2020 at 8:25 AM Tom Kraft > wrote:
>
> Dear statnet,
>
> Thank you for all your wonderful work. I am looking for advice
> on methods applicable to network data that doesn't fit neatly
> into the types of data that regularly?appear in statnet
> tutorials. I have a list of sequential discrete time networks
> involving the same set of actors. At each timepoint, however,
> I only observe some percent of actors in the whole network
> (~30%). For all actors observed I have complete knowledge of
> their ties, including the identity of alters. I would like
> construct a model of these empirical data with relatively
> basic terms that can be used to simulate full temporal
> networks with similar network structures.
>
> Given the temporal nature of the networks, it seems that a
> standard tergm could appropriately be used to model the
> formation and dissolution of ties. Yet given the nature of the
> sampling, for any given network I'm inclined to think that the
> data are essentially egocentric with known alter info because
> if I have information on a node I can be sure that I know all
> the connections of that individual. Thus, I am wondering if it
> is possible to conduct a temporal version of the ergm.ego
> model. From the materials I have found online, it seems this
> should be possible:
>
> "The principles of egocentric inference can be extended to
> temporal ERGMs (TERGMs). While we will not cover that in this
> workshop, an example can be found in another paper that sought
> to evaluate the network hypothsis for racial disparities in
> HIV in the US (Morris et al.?2009)." -
> http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
>
> However, I am unable to find worked examples to follow up on
> that reference/approach and it is not clear to me how to
> implement this. Additionally, it seems like ergm.ego is not
> designed to incorporate information on identifiable alters --
> in which case the methods developed in Kosikinen and Robins
> (2010) perhaps would be more appropriate so that this useful
> information is not ignored.
>
> Alternatively, I could imagine that this analysis is best
> conceived of as a tergm with missing data at each time step.
> In this case info on alter ids could be fully utilized I think.
>
> I would be very grateful if anyone could comment on whether
> one of these approaches seems feasible, or if there are other
> options I might consider. Any materials or references to
> vignettes/tutorials/papers on the topic would also be
> appreciated. Thank you in advance! Best,
>
> Tom
>
> Thomas Kraft
> Postdoctoral Scholar
> Department of Anthropology
> University of California, Santa Barbara
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
>
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
--
------------------------------------------------------------------------
*Interuniversity Centre for Social Science Theory & Methodology*
Department of Sociology, Grote Rozenstraat 31, NL-9712 TG GRONINGEN
steglich.gmw.rug.nl
------------------------------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From morrism at uw.edu Wed May 13 07:43:00 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
In-Reply-To:
References:
Message-ID:
Hi Tom,
The alternative approach may be to treat your observed sample
egocentrically, and use our egocentric inference for ergms/stergms.
If the terms you anticipate using are restricted to those that can be
estimated from egocentric data (basically, the dyad-independence terms
plus degree distributions) these methods would allow you to estimate
stergms, and simulate networks that reproduce the sufficient statistics in
expectation. This is what we are using for our epidemic modeling studies.
Two refs of interest:
Theory:
Krivitsky, P. N., & Morris, M. (2017). Inference for social network models
from egocentrically sampled data, with application to understanding
persistent racial disparities in hiv prevalence in the us. Annals of
Applied Statistics, 11(1), 427-455. doi:10.1214/16-aoas1010
Tutorial for our software:
https://statnet.github.io/Workshops/ergm.ego_tutorial.html
hth,
Martina
On Tue, 12 May 2020, Tom Kraft wrote:
> Hi Zack,
> Thank you for this helpful reference. I had come across this paper in my literature search but should have
> given it more attention.
>
> Is there a place where the source code associated with your paper is available? It remains a bit unclear to me
> how several of the strategies you used (Shat?approximation) might be implemented in R.
>
> I suppose an alternative is to use multiple imputation as discussed here:
> Wang, C., Butts, C. T., Hipp, J. R., Jose, R., & Lakon, C. M. (2016). Multiple imputation for missing edge
> data: a predictive evaluation method with application to add health.?Social networks,?45, 89-98.
>
> Thank you!
> Tom
>
> On Tue, May 12, 2020 at 1:26 PM Zack Almquist wrote:
> Hi Thomas,
> My paper on missing data for TERGM style analysis might be helpful; take a look at it and see if it is
> addressing your problem of interest:
>
> Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis with missing data: theory and
> methods."?Statistica Sinica?28.3 (2018): 1245-1264.
>
> Best,
>
> Zack
> ---
> Zack W. Almquist
> Assistant Professor
> Department of Sociology
> Senior Data Scientist Fellow, eScience Institute
> University of Washington
>
>
> On Tue, May 12, 2020 at 8:25 AM Tom Kraft wrote:
> Dear statnet,
> Thank you for all your wonderful work. I am looking for advice on methods applicable to network
> data that doesn't fit neatly into the types of data that regularly?appear in statnet tutorials. I
> have a list of sequential discrete time networks involving the same set of actors. At each
> timepoint, however, I only observe some percent of actors in the whole network (~30%). For all
> actors observed I have complete knowledge of their ties, including the identity of alters. I would
> like construct a model of these empirical data with relatively basic terms that can be used to
> simulate full temporal networks with similar network structures.
>
> Given the temporal nature of the networks, it seems that a standard tergm could appropriately be
> used to model the formation and dissolution of ties. Yet given the nature of the sampling, for any
> given network I'm inclined to think that the data are essentially egocentric with known alter info
> because if I have information on a node I can be sure that I know all the connections of that
> individual. Thus, I am wondering if it is possible to conduct a temporal version of the ergm.ego
> model. From the materials I have found online, it seems this should be possible:
>
> "The principles of egocentric inference can be extended to temporal ERGMs (TERGMs). While we will
> not cover that in this workshop, an example can be found in another paper that sought to evaluate
> the network hypothsis for racial disparities in HIV in the US (Morris et al.?2009)."
> -?http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
>
> However, I am unable to find worked examples to follow up on that reference/approach and it is not
> clear to me how to implement this. Additionally, it seems like ergm.ego is not designed to
> incorporate information on identifiable alters -- in which case the methods developed in?Kosikinen
> and Robins (2010) perhaps would be more appropriate so that this useful information is not ignored.
>
> Alternatively, I could imagine that this analysis is best conceived of as a tergm with missing data
> at each time step. In this case info on alter ids could be fully utilized I think.
>
> I would be very grateful if anyone could comment on whether one of these approaches seems feasible,
> or if there are other options I might consider. Any materials or references to
> vignettes/tutorials/papers on the topic would also be appreciated. Thank you in advance! Best,
>
> Tom
>
> Thomas Kraft
> Postdoctoral Scholar
> Department of Anthropology
> University of California, Santa Barbara
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
>
>
****************************************************************
Professor Emerita of Sociology and Statistics
Box 354322
University of Washington
Seattle, WA 98195-4322
Office: (206) 685-3402
Dept Office: (206) 543-5882, 543-7237
Fax: (206) 685-7419
morrism@u.washington.edu
http://faculty.washington.edu/morrism/
From kraft.tom at gmail.com Wed May 13 08:45:21 2020
From: kraft.tom at gmail.com (Tom Kraft)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
In-Reply-To:
References:
Message-ID:
Hi Martina,
Thank you for the input! The case I am describing here is in fact part of
an epidemic modeling study. The main problem I see with the approach of
treating the network egocentrically is that I will be forced to sacrifice
information about identifiable alters (again, if a node is observed then I
know with certainty all of the other actors they are linked to and their
identities, and I also know the full list of people they are not connected
to). As such, like you say I won't be able to model dependency terms like
gwesp. But these are undirected contact networks in which if A-B and B-C,
then A-C necessarily exists as well.
To be sure that I am understanding your suggestion: For a given timestep,
completely remove all unobserved nodes from the network. Estimate
parameters using an ego ERGM. Then simulate from the model onto a full
network containing the observed and unobserved nodes and information about
their nodal and edge covariates.
Is this correct? I am familiar with the tutorial link you sent, but it
unfortunately does not cover an egocentric application of a stergm. Do you
know of any worked examples like that which are available?
Thank you again,
Tom
On Wed, May 13, 2020 at 9:43 AM martina morris wrote:
> Hi Tom,
>
> The alternative approach may be to treat your observed sample
> egocentrically, and use our egocentric inference for ergms/stergms.
>
> If the terms you anticipate using are restricted to those that can be
> estimated from egocentric data (basically, the dyad-independence terms
> plus degree distributions) these methods would allow you to estimate
> stergms, and simulate networks that reproduce the sufficient statistics in
> expectation. This is what we are using for our epidemic modeling studies.
>
> Two refs of interest:
>
> Theory:
> Krivitsky, P. N., & Morris, M. (2017). Inference for social network models
> from egocentrically sampled data, with application to understanding
> persistent racial disparities in hiv prevalence in the us. Annals of
> Applied Statistics, 11(1), 427-455. doi:10.1214/16-aoas1010
>
> Tutorial for our software:
> https://statnet.github.io/Workshops/ergm.ego_tutorial.html
>
>
>
>
> hth,
> Martina
>
>
> On Tue, 12 May 2020, Tom Kraft wrote:
>
> > Hi Zack,
> > Thank you for this helpful reference. I had come across this paper in my
> literature search but should have
> > given it more attention.
> >
> > Is there a place where the source code associated with your paper is
> available? It remains a bit unclear to me
> > how several of the strategies you used (Shat approximation) might be
> implemented in R.
> >
> > I suppose an alternative is to use multiple imputation as discussed here:
> > Wang, C., Butts, C. T., Hipp, J. R., Jose, R., & Lakon, C. M. (2016).
> Multiple imputation for missing edge
> > data: a predictive evaluation method with application to add
> health. Social networks, 45, 89-98.
> >
> > Thank you!
> > Tom
> >
> > On Tue, May 12, 2020 at 1:26 PM Zack Almquist wrote:
> > Hi Thomas,
> > My paper on missing data for TERGM style analysis might be helpful; take
> a look at it and see if it is
> > addressing your problem of interest:
> >
> > Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis with
> missing data: theory and
> > methods." Statistica Sinica 28.3 (2018): 1245-1264.
> >
> > Best,
> >
> > Zack
> > ---
> > Zack W. Almquist
> > Assistant Professor
> > Department of Sociology
> > Senior Data Scientist Fellow, eScience Institute
> > University of Washington
> >
> >
> > On Tue, May 12, 2020 at 8:25 AM Tom Kraft wrote:
> > Dear statnet,
> > Thank you for all your wonderful work. I am looking for advice on
> methods applicable to network
> > data that doesn't fit neatly into the types of data that
> regularly appear in statnet tutorials. I
> > have a list of sequential discrete time networks involving the same set
> of actors. At each
> > timepoint, however, I only observe some percent of actors in the whole
> network (~30%). For all
> > actors observed I have complete knowledge of their ties, including the
> identity of alters. I would
> > like construct a model of these empirical data with relatively basic
> terms that can be used to
> > simulate full temporal networks with similar network structures.
> >
> > Given the temporal nature of the networks, it seems that a standard
> tergm could appropriately be
> > used to model the formation and dissolution of ties. Yet given the
> nature of the sampling, for any
> > given network I'm inclined to think that the data are essentially
> egocentric with known alter info
> > because if I have information on a node I can be sure that I know all
> the connections of that
> > individual. Thus, I am wondering if it is possible to conduct a temporal
> version of the ergm.ego
> > model. From the materials I have found online, it seems this should be
> possible:
> >
> > "The principles of egocentric inference can be extended to temporal
> ERGMs (TERGMs). While we will
> > not cover that in this workshop, an example can be found in another
> paper that sought to evaluate
> > the network hypothsis for racial disparities in HIV in the US (Morris et
> al. 2009)."
> > - http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
> >
> > However, I am unable to find worked examples to follow up on that
> reference/approach and it is not
> > clear to me how to implement this. Additionally, it seems like ergm.ego
> is not designed to
> > incorporate information on identifiable alters -- in which case the
> methods developed in Kosikinen
> > and Robins (2010) perhaps would be more appropriate so that this useful
> information is not ignored.
> >
> > Alternatively, I could imagine that this analysis is best conceived of
> as a tergm with missing data
> > at each time step. In this case info on alter ids could be fully
> utilized I think.
> >
> > I would be very grateful if anyone could comment on whether one of these
> approaches seems feasible,
> > or if there are other options I might consider. Any materials or
> references to
> > vignettes/tutorials/papers on the topic would also be appreciated. Thank
> you in advance! Best,
> >
> > Tom
> >
> > Thomas Kraft
> > Postdoctoral Scholar
> > Department of Anthropology
> > University of California, Santa Barbara
> > _______________________________________________
> > statnet_help mailing list
> > statnet_help@u.washington.edu
> > http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
> >
> >
> >
>
> ****************************************************************
> Professor Emerita of Sociology and Statistics
> Box 354322
> University of Washington
> Seattle, WA 98195-4322
>
> Office: (206) 685-3402
> Dept Office: (206) 543-5882, 543-7237
> Fax: (206) 685-7419
>
> morrism@u.washington.edu
> http://faculty.washington.edu/morrism/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From zalmquist at uw.edu Wed May 13 09:22:16 2020
From: zalmquist at uw.edu (Zack Almquist)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
In-Reply-To:
References:
Message-ID:
Hi Tom,
I really like Martina's proposal as I think it makes a lot of sense given
the description of your data!
There are obviously a lot of dynamic network models and ways for
handling missing data -- I trust you to find the best one for your problem!
Christian has provided a nice set of papers for the SAOM framework that has
a rich history in dynamic network modeling (as he pointed out); and there
are also the relational event modeling frameworks (if you are going to work
in continuous time, e.g., Butts, Carter T. "4. A Relational Event Framework
for Social Action." *Sociological Methodology* 38.1 (2008): 155-200). Note
I am not trying to be comprehensive!
As per my method, I have a paper in JCGS with some software:
Mallik, A., & Almquist, Z. W. (2019). Stable Multiple Time Step
Simulation/Prediction From Lagged Dynamic Network Regression Models. *Journal
of Computational and Graphical Statistics*, *28*(4), 967-979.
https://cran.r-project.org/web/packages/dnr/index.html
That could be readily adapted for the methods in
Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis with
missing data: theory and methods." *Statistica Sinica* 28.3 (2018):
1245-1264.
Best,
Zack
---
Zack W. Almquist
Assistant Professor
Department of Sociology
Senior Data Scientist Fellow, eScience Institute
University of Washington
On Wed, May 13, 2020 at 8:47 AM Tom Kraft wrote:
> Hi Martina,
>
> Thank you for the input! The case I am describing here is in fact part of
> an epidemic modeling study. The main problem I see with the approach of
> treating the network egocentrically is that I will be forced to sacrifice
> information about identifiable alters (again, if a node is observed then I
> know with certainty all of the other actors they are linked to and their
> identities, and I also know the full list of people they are not connected
> to). As such, like you say I won't be able to model dependency terms like
> gwesp. But these are undirected contact networks in which if A-B and B-C,
> then A-C necessarily exists as well.
>
> To be sure that I am understanding your suggestion: For a given timestep,
> completely remove all unobserved nodes from the network. Estimate
> parameters using an ego ERGM. Then simulate from the model onto a full
> network containing the observed and unobserved nodes and information about
> their nodal and edge covariates.
>
> Is this correct? I am familiar with the tutorial link you sent, but it
> unfortunately does not cover an egocentric application of a stergm. Do you
> know of any worked examples like that which are available?
>
> Thank you again,
> Tom
>
> On Wed, May 13, 2020 at 9:43 AM martina morris wrote:
>
>> Hi Tom,
>>
>> The alternative approach may be to treat your observed sample
>> egocentrically, and use our egocentric inference for ergms/stergms.
>>
>> If the terms you anticipate using are restricted to those that can be
>> estimated from egocentric data (basically, the dyad-independence terms
>> plus degree distributions) these methods would allow you to estimate
>> stergms, and simulate networks that reproduce the sufficient statistics in
>> expectation. This is what we are using for our epidemic modeling studies.
>>
>> Two refs of interest:
>>
>> Theory:
>> Krivitsky, P. N., & Morris, M. (2017). Inference for social network
>> models
>> from egocentrically sampled data, with application to understanding
>> persistent racial disparities in hiv prevalence in the us. Annals of
>> Applied Statistics, 11(1), 427-455. doi:10.1214/16-aoas1010
>>
>> Tutorial for our software:
>> https://statnet.github.io/Workshops/ergm.ego_tutorial.html
>>
>>
>>
>>
>> hth,
>> Martina
>>
>>
>> On Tue, 12 May 2020, Tom Kraft wrote:
>>
>> > Hi Zack,
>> > Thank you for this helpful reference. I had come across this paper in
>> my literature search but should have
>> > given it more attention.
>> >
>> > Is there a place where the source code associated with your paper is
>> available? It remains a bit unclear to me
>> > how several of the strategies you used (Shat approximation) might be
>> implemented in R.
>> >
>> > I suppose an alternative is to use multiple imputation as discussed
>> here:
>> > Wang, C., Butts, C. T., Hipp, J. R., Jose, R., & Lakon, C. M. (2016).
>> Multiple imputation for missing edge
>> > data: a predictive evaluation method with application to add
>> health. Social networks, 45, 89-98.
>> >
>> > Thank you!
>> > Tom
>> >
>> > On Tue, May 12, 2020 at 1:26 PM Zack Almquist wrote:
>> > Hi Thomas,
>> > My paper on missing data for TERGM style analysis might be helpful;
>> take a look at it and see if it is
>> > addressing your problem of interest:
>> >
>> > Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis with
>> missing data: theory and
>> > methods." Statistica Sinica 28.3 (2018): 1245-1264.
>> >
>> > Best,
>> >
>> > Zack
>> > ---
>> > Zack W. Almquist
>> > Assistant Professor
>> > Department of Sociology
>> > Senior Data Scientist Fellow, eScience Institute
>> > University of Washington
>> >
>> >
>> > On Tue, May 12, 2020 at 8:25 AM Tom Kraft wrote:
>> > Dear statnet,
>> > Thank you for all your wonderful work. I am looking for advice on
>> methods applicable to network
>> > data that doesn't fit neatly into the types of data that
>> regularly appear in statnet tutorials. I
>> > have a list of sequential discrete time networks involving the same set
>> of actors. At each
>> > timepoint, however, I only observe some percent of actors in the whole
>> network (~30%). For all
>> > actors observed I have complete knowledge of their ties, including the
>> identity of alters. I would
>> > like construct a model of these empirical data with relatively basic
>> terms that can be used to
>> > simulate full temporal networks with similar network structures.
>> >
>> > Given the temporal nature of the networks, it seems that a standard
>> tergm could appropriately be
>> > used to model the formation and dissolution of ties. Yet given the
>> nature of the sampling, for any
>> > given network I'm inclined to think that the data are essentially
>> egocentric with known alter info
>> > because if I have information on a node I can be sure that I know all
>> the connections of that
>> > individual. Thus, I am wondering if it is possible to conduct a
>> temporal version of the ergm.ego
>> > model. From the materials I have found online, it seems this should be
>> possible:
>> >
>> > "The principles of egocentric inference can be extended to temporal
>> ERGMs (TERGMs). While we will
>> > not cover that in this workshop, an example can be found in another
>> paper that sought to evaluate
>> > the network hypothsis for racial disparities in HIV in the US (Morris
>> et al. 2009)."
>> > -
>> http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
>> >
>> > However, I am unable to find worked examples to follow up on that
>> reference/approach and it is not
>> > clear to me how to implement this. Additionally, it seems like ergm.ego
>> is not designed to
>> > incorporate information on identifiable alters -- in which case the
>> methods developed in Kosikinen
>> > and Robins (2010) perhaps would be more appropriate so that this useful
>> information is not ignored.
>> >
>> > Alternatively, I could imagine that this analysis is best conceived of
>> as a tergm with missing data
>> > at each time step. In this case info on alter ids could be fully
>> utilized I think.
>> >
>> > I would be very grateful if anyone could comment on whether one of
>> these approaches seems feasible,
>> > or if there are other options I might consider. Any materials or
>> references to
>> > vignettes/tutorials/papers on the topic would also be appreciated.
>> Thank you in advance! Best,
>> >
>> > Tom
>> >
>> > Thomas Kraft
>> > Postdoctoral Scholar
>> > Department of Anthropology
>> > University of California, Santa Barbara
>> > _______________________________________________
>> > statnet_help mailing list
>> > statnet_help@u.washington.edu
>> > http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>> >
>> >
>> >
>>
>> ****************************************************************
>> Professor Emerita of Sociology and Statistics
>> Box 354322
>> University of Washington
>> Seattle, WA 98195-4322
>>
>> Office: (206) 685-3402
>> Dept Office: (206) 543-5882, 543-7237
>> Fax: (206) 685-7419
>>
>> morrism@u.washington.edu
>> http://faculty.washington.edu/morrism/
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From morrism at uw.edu Wed May 13 10:27:46 2020
From: morrism at uw.edu (martina morris)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] temporal ergm with partially observed networks
In-Reply-To:
References:
Message-ID:
Hi Tom,
Identifiable alters are only important if:
1. you are interested in a fixed network on this particular set of
nodes, or
2. the structures you want to model are higher order than dyadic
independent/degree based configurations.
Even in case (2), it may be that these lower order processes place such
constraints on the higher order stats (e.g., the geodesic distn), that you
can reproduce the higher order stats by modeling the lower order ones.
And in your case, you might be able to test that, by modeling the network
egocentrically, simulating from it, and comparing some of the higher order
stats for the observed and simulated data. (with the residual caveat that
the missing data must be accounted for somehow)
You should take a look at our Network Modeling for Epidemics course
materials. We teach that workshop each year. This year there will be a
remote option (in person only if possible).
Course materials/info: http://statnet.org/nme/
Applications still open for this year's course in August, link on the site
above.
hth,
mm
On Wed, 13 May 2020, Tom Kraft wrote:
> Hi Martina,
> Thank you for the input! The case I am describing here is in fact part of an epidemic modeling study. The main
> problem I see with the approach of treating the network egocentrically is that I will be forced to sacrifice
> information about identifiable alters (again, if a node is observed then I know with certainty all of the other
> actors they are linked to and their identities, and I also know the full list of people they are not connected
> to). As such, like you say I won't be able to model dependency terms like gwesp. But these are undirected
> contact networks in which if A-B and B-C, then A-C necessarily exists as well.
>
> To be sure that I am understanding your suggestion: For a given timestep, completely remove all unobserved
> nodes from the network. Estimate parameters using an ego ERGM. Then simulate from the model onto a full network
> containing the observed and unobserved nodes and information about their nodal and edge covariates.
>
> Is this correct? I am familiar with the tutorial link you sent, but it unfortunately does not cover an
> egocentric application of a stergm. Do you know of any worked examples like that which are available?
>
> Thank you again,
> Tom
>
> On Wed, May 13, 2020 at 9:43 AM martina morris wrote:
> Hi Tom,
>
> The alternative approach may be to treat your observed sample
> egocentrically, and use our egocentric inference for ergms/stergms.
>
> If the terms you anticipate using are restricted to those that can be
> estimated from egocentric data (basically, the dyad-independence terms
> plus degree distributions) these methods would allow you to estimate
> stergms, and simulate networks that reproduce the sufficient statistics in
> expectation.? This is what we are using for our epidemic modeling studies.
>
> Two refs of interest:
>
> Theory:
> Krivitsky, P. N., & Morris, M. (2017). Inference for social network models
> from egocentrically sampled data, with application to understanding
> persistent racial disparities in hiv prevalence in the us. Annals of
> Applied Statistics, 11(1), 427-455. doi:10.1214/16-aoas1010
>
> Tutorial for our software:
> https://statnet.github.io/Workshops/ergm.ego_tutorial.html
>
>
>
>
> hth,
> Martina
>
>
> On Tue, 12 May 2020, Tom Kraft wrote:
>
> > Hi Zack,
> > Thank you for this helpful reference. I had come across this paper in my literature search but
> should have
> > given it more attention.
> >
> > Is there a place where the source code associated with your paper is available? It remains a bit
> unclear to me
> > how several of the strategies you used (Shat?approximation) might be implemented in R.
> >
> > I suppose an alternative is to use multiple imputation as discussed here:
> > Wang, C., Butts, C. T., Hipp, J. R., Jose, R., & Lakon, C. M. (2016). Multiple imputation for
> missing edge
> > data: a predictive evaluation method with application to add health.?Social networks,?45, 89-98.
> >
> > Thank you!
> > Tom
> >
> > On Tue, May 12, 2020 at 1:26 PM Zack Almquist wrote:
> >? ? ? ?Hi Thomas,
> > My paper on missing data for TERGM style analysis might be helpful; take a look at it and see if
> it is
> > addressing your problem of interest:
> >
> > Almquist, Zack W., and Carter T. Butts. "Dynamic network analysis with missing data: theory and
> > methods."?Statistica Sinica?28.3 (2018): 1245-1264.
> >
> > Best,
> >
> > Zack
> > ---
> > Zack W. Almquist
> > Assistant Professor
> > Department of Sociology
> > Senior Data Scientist Fellow, eScience Institute
> > University of Washington
> >
> >
> > On Tue, May 12, 2020 at 8:25 AM Tom Kraft wrote:
> >? ? ? ?Dear statnet,
> > Thank you for all your wonderful work. I am looking for advice on methods applicable to network
> > data that doesn't fit neatly into the types of data that regularly?appear in statnet tutorials. I
> > have a list of sequential discrete time networks involving the same set of actors. At each
> > timepoint, however, I only observe some percent of actors in the whole network (~30%). For all
> > actors observed I have complete knowledge of their ties, including the identity of alters. I
> would
> > like construct a model of these empirical data with relatively basic terms that can be used to
> > simulate full temporal networks with similar network structures.
> >
> > Given the temporal nature of the networks, it seems that a standard tergm could appropriately be
> > used to model the formation and dissolution of ties. Yet given the nature of the sampling, for
> any
> > given network I'm inclined to think that the data are essentially egocentric with known alter
> info
> > because if I have information on a node I can be sure that I know all the connections of that
> > individual. Thus, I am wondering if it is possible to conduct a temporal version of the ergm.ego
> > model. From the materials I have found online, it seems this should be possible:
> >
> > "The principles of egocentric inference can be extended to temporal ERGMs (TERGMs). While we will
> > not cover that in this workshop, an example can be found in another paper that sought to evaluate
> > the network hypothsis for racial disparities in HIV in the US (Morris et al.?2009)."
> > -?http://statnet.org/Workshops/ergm.ego_tutorial.html#6_example_analysis
> >
> > However, I am unable to find worked examples to follow up on that reference/approach and it is
> not
> > clear to me how to implement this. Additionally, it seems like ergm.ego is not designed to
> > incorporate information on identifiable alters -- in which case the methods developed
> in?Kosikinen
> > and Robins (2010) perhaps would be more appropriate so that this useful information is not
> ignored.
> >
> > Alternatively, I could imagine that this analysis is best conceived of as a tergm with missing
> data
> > at each time step. In this case info on alter ids could be fully utilized I think.
> >
> > I would be very grateful if anyone could comment on whether one of these approaches seems
> feasible,
> > or if there are other options I might consider. Any materials or references to
> > vignettes/tutorials/papers on the topic would also be appreciated. Thank you in advance! Best,
> >
> > Tom
> >
> > Thomas Kraft
> > Postdoctoral Scholar
> > Department of Anthropology
> > University of California, Santa Barbara
> > _______________________________________________
> > statnet_help mailing list
> > statnet_help@u.washington.edu
> > http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
> >
> >
> >
>
> ****************************************************************
> ? Professor Emerita of Sociology and Statistics
> ? Box 354322
> ? University of Washington
> ? Seattle, WA 98195-4322
>
> ? Office:? ? ? ? (206) 685-3402
> ? Dept Office:? ?(206) 543-5882, 543-7237
> ? Fax:? ? ? ? ? ?(206) 685-7419
>
> morrism@u.washington.edu
> http://faculty.washington.edu/morrism/
>
>
>
****************************************************************
Professor Emerita of Sociology and Statistics
Box 354322
University of Washington
Seattle, WA 98195-4322
Office: (206) 685-3402
Dept Office: (206) 543-5882, 543-7237
Fax: (206) 685-7419
morrism@u.washington.edu
http://faculty.washington.edu/morrism/
From barnaba at umcs.edu.pl Sat May 16 09:42:19 2020
From: barnaba at umcs.edu.pl (Barnaba Danieluk)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] ERGM and statistical power (and effect size)
Message-ID: <55d8bd4f71b9abbe293c25dc503c0b7e@umcs.edu.pl>
Dear Statnet,
One of the reviewers asked me about statistical power of my TERGM model.
Is there any parameter in ERGM (TERGM) that I can use as a effect size
indicator? Is it something similar to R-square in ERGM? And how can I
count the statistical power of my analysis?
I would be very grateful for every advice. Thank you in advance
Best regards,
Barnaba
--
Barnaba Danieluk
Assistant Professor
Institute of Psychology
Maria Curie-Sklodowska University
Lublin, Poland
From barnaba at umcs.edu.pl Sat May 16 09:42:19 2020
From: barnaba at umcs.edu.pl (Barnaba Danieluk)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] ERGM and statistical power (and effect size)
Message-ID: <55d8bd4f71b9abbe293c25dc503c0b7e@umcs.edu.pl>
Dear Statnet,
One of the reviewers asked me about statistical power of my TERGM model.
Is there any parameter in ERGM (TERGM) that I can use as a effect size
indicator? Is it something similar to R-square in ERGM? And how can I
count the statistical power of my analysis?
I would be very grateful for every advice. Thank you in advance
Best regards,
Barnaba
--
Barnaba Danieluk
Assistant Professor
Institute of Psychology
Maria Curie-Sklodowska University
Lublin, Poland
From buttsc at uci.edu Sat May 16 16:19:40 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] ERGM and statistical power (and effect size)
In-Reply-To: <55d8bd4f71b9abbe293c25dc503c0b7e@umcs.edu.pl>
References: <55d8bd4f71b9abbe293c25dc503c0b7e@umcs.edu.pl>
Message-ID:
Hi, Barnaba -
This depends on what you mean by "effect size" (and, for that matter,
the notion of power that you want to use).? Your coefficients may be
directly interpreted in terms of their effect on the conditional
log-odds of a tie being present, and this is the most immediate and
natural notion of "effect size" in a TERGM.? One can also talk about
"effect sizes" in terms of quantities such as e.g. the difference in
conditional expectations of selected network properties with a
coefficient at its estimated value versus the coefficient set to zero (a
"knock-out" study, if you will), with these quantities being
approximated by simulation.? This can be a very useful and insightful
approach, but it requires you to be specific about what, precisely, you
want to know about the relationship between the coefficient of interest
and model behavior.? There is no "one size fits all" approach to this
family of questions; often, however, your substantive problem will
strongly motivate a particular choice of outcome, and looking at how
this varies as a function of your parameter of interest can give you a
lot of insight.
(For example, let's say that you are interested in the amount of mixing
between two groups, as assessed e.g. by a nodemix statistic.? You fit a
TERGM with a nodemix effect (among other things), and obtain an
associated parameter estimate.? What is the marginal impact of that
parameter on between-group mixing?? One way to probe that question is to
simulate draws from the fitted model, and from the same model with the
nodemix parameter set to zero, comparing the mean nodemix statistic in
the two cases.? The difference tells you how much more or less mixing
you would expect to obtain, with all other social forces held constant,
if the specific force parameterized by your nodemix term were not
active.? This is only one of many comparisons that one can perform, but
it illustrates the concept.)
With respect to power, this always becomes complex as soon as one leaves
the world of simple null hypothesis tests.? As a purely practical
matter, you may get more mileage out of answering a closely related but
distinct question: what hypotheses regarding your parameters can you
/not/ reject, given the data?? This amounts to looking at your
confidence intervals.? E.g., if you have a nodematch effect for
membership in group A with a point estimate of 1 and a 95% CI of 0.5 to
1.5, then an associated null hypothesis test would reject the hypothesis
(and the 0.05 level) that the conditional log-odds of an i,j edge are
increased by less than 0.5 when? i and j both belong to A, or likewise
that the conditional log-odds of an i,j edge are increased by more than
1.5 in the same circumstance.? You can thus /exclude/ (in a Frequentist
sense) effect sizes (on log-odds scale) smaller than 0.5, or larger than
1.5.? If you have very little power for estimating an effect, you will
generally find that the range of values that cannot be excluded is very
large (i.e., the CIs are wide); by turns, if your CIs are small, then
this is telling you that you have enough precision to be able to exclude
all values (at least in terms of rejection of the associated null
hypothesis test) outside of a very narrow range.? If your reviewer is
not asking about power merely for the satisfaction of asking about it,
then you may well be able to get at their substantive concern by a
closer interpretation of your confidence intervals.? (This is especially
true for non-significant results, where you may not be able to determine
the sign of an effect, but may still be able to reject the hypothesis
that the effect is large.? From a substantive standpoint, this often
enough to falsify a theory.)
Finally, it is possible to define R^2-like measures for TERGMs et al.,
but to date I don't know that anyone has found such things to be very
helpful.? In particular, you can use any of the many deviance reduction
indices (aka "deviance R^2" measures) for an ERGM or a TERGM, as one
would with any other model with a well-defined deviance.? The reduction
in deviance from the null deviance is certainly associated with improved
fit (and is of course the basis for the AIC et al.), but my sense is
that measures like 1-(residual deviance)/(null deviance) are only weakly
and heuristically related to how well the model works in practice.?
Still, if your reviewer wants a deviance R^2, it's an easy and probably
harmless thing to provide.
These are complex questions, but hopefully the above is useful!
-Carter
On 5/16/20 9:42 AM, Barnaba Danieluk wrote:
> Dear Statnet,
>
> One of the reviewers asked me about statistical power of my TERGM
> model. Is there any parameter in ERGM (TERGM) that I can use as a
> effect size indicator? Is it something similar to R-square in ERGM?
> And how can I count the statistical power of my analysis?
>
> I would be very grateful for every advice. Thank you in advance
>
> Best regards,
> Barnaba
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From lingchuding at gmail.com Sat Jun 6 07:47:21 2020
From: lingchuding at gmail.com (CHU-DING LING)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Help seeking on calculating betweenness centrality
with valued ties
Message-ID:
Dear all,
I am trying to calculate the betweenness centrality with valued ties, but I
do not figure it out. Here are the dataset and codes.
1. This is an example dataset with only four nodes/individuals. When
collecting the network data, I asked the participants to answer the
question about friendship tie on a 7-point Likert scale. So, this is a
*directed* and *valued* network. Moreover, when inputting the network data,
I adopted the edge list format and saved it into a CSV file. The details of
the data are as follows:
Actor Target Friend
1001 1002 5
1001 1003 6
1001 1004 5
1002 1001 6
1002 1003 6
1002 1004 6
1003 1001 4
1003 1002 4
1003 1004 4
1004 1001 6
1004 1002 6
1004 1003 6
2. Then I ran the following codes to calculate the betweenness centrality:
library(statnet)
#Step 1. read the edgelist format dataset into R
Mydata <- read.table("Example.csv", header=TRUE, sep=",")
#Step 2. convert an edgelist matrix with valued edges/ties into a network
Mynet <- network (Mydata)
#Step 3. calculate betweenness centrality but fail to account for the
value/weight of the tie
betweenness (Mynet)
3. The results came out are as follows:
[1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0
[43] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[85] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[127] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[169] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[211] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[253] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[295] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[337] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[379] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[421] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[463] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[505] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[547] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[589] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[631] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[673] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[715] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[757] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[799] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[841] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[883] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[925] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
[967] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
[ reached getOption("max.print") -- omitted 4 entries ]
It seems that the package treated the IDs of actor and target as edges when
computing. *If I want to keep the numeric IDs for the further merging with
other variables, what can I do to solve this problem?*
Also, if I replace the numeric IDs with strings and organize the data as
follows:
Actor Target Friend
A B 5
A C 6
A D 5
B A 6
B C 6
B D 6
C A 4
C B 4
C D 4
D A 6
D B 6
D C 6
Then, I re-ran the Step 2 and Step 3, the results were as follows:
[1] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 44.0 44.0 44.0
I think the results are also wrong since the expected results should be
four values. However, there are 15 values with the first 12 looking equal.
I have searched archival of the list, but I failed to locate the
information that can completely solve my problems. So, I am wondering
whether any colleagues here could share with me any information about this.
I would be grateful if you can provide me any suggestions or references.
Many thanks in advance!
Best,
Chuding
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From dluke at wustl.edu Sat Jun 6 09:08:28 2020
From: dluke at wustl.edu (Luke, Douglas)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] statnet_help Digest, Vol 166, Issue 1
In-Reply-To:
References:
Message-ID:
Chuding,
One quick tip: you need to make sure that you are actually creating a real network object before you calculate appropriate network statistics.
The first missing step in your code is you need to tell the network creation function that you are working with an edgelist.
This should work:
net1 <- as.network(net1.df, matrix.type="edgelist")
summary(net1)
Where 'net1.df' is your dataframe based on the CSV file, and 'net1' will be your new, directed network object.
However, this will not read in the Friend valued ties, that will require more work.
I would encourage you to understand what the betweenness centrality scores are telling you with the non-valued ties first. I myself find it much harder to interpret betweenness of valued-ties.
Hope this helps,
--Doug--
Douglas Luke
Director, Center for Public Health Systems Science
Director, Doctoral Program in Public Health Sciences
Professor, Brown School
Washington University in St. Louis
Campus Box 1196, One Brookings Drive
St. Louis, MO 63130
email: dluke@wustl.edu
Website: http://cphss.wustl.edu
________________________________
From: statnet_help on behalf of statnet_help-request@mailman13.u.washington.edu
Sent: Saturday, June 6, 2020 9:48 AM
To: statnet_help@u.washington.edu
Subject: statnet_help Digest, Vol 166, Issue 1
Send statnet_help mailing list submissions to
statnet_help@u.washington.edu
To subscribe or unsubscribe via the World Wide Web, visit
http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
or, via email, send a message with subject or body 'help' to
statnet_help-request@mailman13.u.washington.edu
You can reach the person managing the list at
statnet_help-owner@mailman13.u.washington.edu
When replying, please edit your Subject line so it is more specific
than "Re: Contents of statnet_help digest..."
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From buttsc at uci.edu Sat Jun 6 14:52:55 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Help seeking on calculating betweenness
centrality with valued ties
In-Reply-To:
References:
Message-ID: <31086dbb-2bea-b400-198c-f178d9094232@uci.edu>
Hi, Chuding -
You may find the valued network tutorial to be helpful here:
http://statnet.org/Workshops/valued.html
If you are interested in calculating betweenness per se, however, it is
probably most straightforward to either store your data in either a
valued adjacency matrix, or an sna edgelist (see ?as.edgelist.sna).?
When you pass the data object to the betweenness function, you will also
need to tell it to use the edge values, since it otherwise will assume
that you want to calculate betweenness on the underlying unvalued
digraph (see ?betweenness).
Note that directly passing a network object that happens to contain edge
values to betweenness et al. will not result in those values being used;
the reason is that a network object can contain any number of different
sets of edge variables (which need not even be "values" as such - you
can have a dozen network objects attached to each edge, if you like),
and the NLI functions have no way of knowing which if any values you
want it to use.? Thus, the way to do it is to pass an object with the
values extracted (e.g., a valued sociomatrix or sna edgelist, as noted
above).? If you have encoded your edge values into a network object,
this can easily be done with code like so:
library(sna)
data(emon)
betweenness(as.edgelist.sna(emon[[1]],"Frequency"), ignore.eval=FALSE)
In this case, "Frequency" is a numeric variable associated with the
edges in emon[[1]].? If we left ignore.eval at its default value of
TRUE, we would obtain the corresponding scores for the underlying
(unvalued) network.? Likewise, if we just called betweenness(emon[[1]])
- with or without ignore.eval=FALSE - the betweenness function would
have no idea that we wanted the "Frequency" variable, and would hence
operate on the raw graph structure.? This same approach is used by most
of the functional indices in statnet, but you should always check the
individual help pages for details; some indices are not defined for
valued data, and others may have additional options that are important
to consider.
In line with that last note - and mostly for the benefit of others who
may encounter the thread - I would also observe that generalizing node
or graph-level indices to the valued case is not always trivial, and in
some cases the "valued" version of an index may only make sense when the
edge values are interpreted in a particular way.? For instance, indices
based on geodesics (including betweenness) implicitly treat edge values
as distances, with the additional implicit assumption that a length of
an i,j path is the sum of the values of the edges contained within it.
If one's data were to consist e.g. of subjective ratings of positive
feeling (e.g., how much does ego "like" alter on some scale or other),
then using these values in this way would make little sense; statnet
would dutifully compute the indices for you in such a situation if you
asked it to, but they would have no obvious substantive interpretation.?
In some cases, it may be possible to transform one's original
observations into a form that is more appropriate for such an analysis
(e.g., transforming similarity scores to distance scores), but in other
cases there may be a mismatch between the data properties tacitly
assumed by the indices (at least, in their conventional interpretation*)
and what was actually measured.? In such cases, it may be wise to
consider a different index that is more appropriate to one's problem.?
Finally, it can also be important to ensure that valued data is coded in
the way that one intends.? For instance, the Drabek et al. EMON data
used above codes interaction frequency from 1 to 4, with 1 being the
highest level of frequency ("continuous") and four being the lowest
non-zero value ("about once a day or less").? For betweenness or
closeness (where we are treating the edges as if they are distances),
that at least makes ordinal sense, but calculating valued degree without
recoding would give us rather perverse results.? (Whether it is
reasonable to treat path lengths as additive with respect to such a
measure is an interesting question that I will not answer, but instead
allow to hang in the proverbial air for silent contemplation.)? So
anyway, be sure to verify that your index is using your edge values in a
way that both makes substantive sense and that is compatible with how
they are coded.
Hope that helps,
-Carter
* Technically speaking, an index doesn't "assume" anything.? It is
simply a function of the network.? But our conventional
/interpretations/ of what structural indices tell us about social
networks usually involve quite a few assumptions, often tacit.? When
working with valued data, it is often necessary to become more explicit
about what one is assuming.
On 6/6/20 7:47 AM, CHU-DING LING wrote:
>
> Dear all,
>
> I am trying to calculate the betweenness centrality with valued ties,
> but I do not figure it out. Here are the dataset and codes.
>
> 1. This is an example dataset with only four nodes/individuals. When
> collecting the network data, I asked the participants to answer the
> question about friendship tie on a 7-point Likert scale. So, this is a
> *directed* and *valued* network. Moreover, when inputting the network
> data, I adopted the edge list format and saved it into a CSV file. The
> details of the data are as follows:
>
> Actor Target?? Friend
>
> 1001? 1002? 5
>
> 1001? 1003? 6
>
> 1001? 1004? 5
>
> 1002? 1001? 6
>
> 1002? 1003? 6
>
> 1002? 1004? 6
>
> 1003? 1001? 4
>
> 1003? 1002? 4
>
> 1003? 1004? 4
>
> 1004? 1001? 6
>
> 1004? 1002? 6
>
> 1004? 1003? 6
>
> 2. Then I ran the following codes to calculate the betweenness centrality:
>
> library(statnet)
>
> #Step 1. read the edgelist format dataset into R
>
> Mydata <- read.table("Example.csv", header=TRUE, sep=",")
>
> #Step 2. convert an edgelist matrix with valued edges/ties into a network
>
> Mynet <- network (Mydata)
>
> #Step 3. calculate betweenness centrality but fail to account for the
> value/weight of the tie
>
> betweenness (Mynet)
>
> 3. The results came out are as follows:
>
> [1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0
>
> [43] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0
>
> [85] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0
>
> [127] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [169] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [211] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [253] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [295] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [337] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [379] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [421] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [463] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [505] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [547] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [589] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [631] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [673] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [715] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [757] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [799] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [841] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [883] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [925] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0 0 0
>
> [967] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>
> [ reached getOption("max.print") -- omitted 4 entries ]
>
> It seems that the package treated the IDs of actor and target as edges
> when computing. *If I want to keep the numeric IDs for the further
> merging with other variables, what can I do to solve this problem?*
>
> Also, if I replace the numeric IDs with strings and organize the data
> as follows:
>
> Actor Target Friend
>
> A? B? 5
>
> A? C? 6
>
> A? D? 5
>
> B? A? 6
>
> B? C? 6
>
> B? D? 6
>
> C? A? 4
>
> C? B? 4
>
> C? D? 4
>
> D? A? 6
>
> D? B? 6
>
> D? C? 6
>
> Then, I re-ran the Step 2 and Step 3, the results were as follows:
>
> [1] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 44.0 44.0 44.0
>
> I think the results are also wrong since the expected results should
> be four values. However, there are 15 values with the first 12 looking
> equal.
>
> I have searched archival of the list, but I failed to locate the
> information that can completely solve my problems. So, I am wondering
> whether any colleagues here could share with me any information about
> this. I would be grateful if you can provide me any suggestions or
> references. Many thanks in advance!
>
> Best,
>
> Chuding
>
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From lingchuding at gmail.com Sat Jun 6 20:27:16 2020
From: lingchuding at gmail.com (CHU-DING LING)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Help seeking on calculating betweenness
centrality with valued ties
In-Reply-To:
References:
<31086dbb-2bea-b400-198c-f178d9094232@uci.edu>
Message-ID:
Another problem is that even though I transform those valued ties into
binary ties, the package would still treat everyone in the network is
connected with the others, as long as I have 12 edges in the 4-vertex
graph. I tried to remove the cases with a value of edge as 0 in the graph
and calculated the indegree of each vertex, the results are consistent with
the reality in the network. So, it seems that the package would always
treat the column regarding the value of the edges as a weight no matter it
is a binary or continuous variable. Am I right?
Best,
Chuding
CHU-DING LING ?2020?6?7??? ??9:44???
> Dear Carter,
>
>
>
> Thanks for your prompt reply. Actually, I have noticed the tutorial on
> valued ties. However, I did not follow many steps of it. And I failed to
> reproduce the results by just copying and running the codes in the
> examples. There are some errors. That is why I have to post this question
> in the list.
>
>
>
> After reading your suggestions, I have attempted to convert the network
> object into a matrix:
>
>
>
> Mynet.matrix = as.matrix(Mynet)
>
> class (Mynet.matrix)
>
> [1] "matrix" "array"
>
>
>
> And I ran the code to calculate the betweenness centrality again:
>
>
>
> betweenness (as.edgelist.sna (Mynet.matrix[[1]],"Friend"),
> ignore.eval=FALSE)
>
> Error in as.edgelist.sna (Mynet.matrix[[1]], "Friend"):
>
> as.edgelist.sna input must be an adjacency matrix/array, edgelist
> matrix, network, or sparse matrix, or list thereof.
>
>
>
> According to the results from the ?class (Mynet.matrix)? command, I think
> the ?Mynet.matrix? object is a matrix now. *Why does the ?as.edgelist.sna
> ()? cannot work it out?*
>
>
>
> Also, I have explored the class of the ?emon? object in your example and
> results showed it is a list.
>
>
>
> class(emon)
>
> [1] "list"
>
>
>
> *Does this difference lead the ?as.edgelist.sna ()? command to reporting
> the errors?*
>
>
>
> I am sorry to bother you again. I am a novice of R and ?statnet?, so I am
> wondering whether you can give me more detailed illustration based on my
> example dataset. Many thanks in advance!
>
>
>
> Best,
>
> Chuding
>
>
> Carter T. Butts ?2020?6?7??? ??5:54???
>
>> Hi, Chuding -
>>
>> You may find the valued network tutorial to be helpful here:
>> http://statnet.org/Workshops/valued.html
>>
>> If you are interested in calculating betweenness per se, however, it is
>> probably most straightforward to either store your data in either a valued
>> adjacency matrix, or an sna edgelist (see ?as.edgelist.sna). When you pass
>> the data object to the betweenness function, you will also need to tell it
>> to use the edge values, since it otherwise will assume that you want to
>> calculate betweenness on the underlying unvalued digraph (see
>> ?betweenness).
>>
>> Note that directly passing a network object that happens to contain edge
>> values to betweenness et al. will not result in those values being used;
>> the reason is that a network object can contain any number of different
>> sets of edge variables (which need not even be "values" as such - you can
>> have a dozen network objects attached to each edge, if you like), and the
>> NLI functions have no way of knowing which if any values you want it to
>> use. Thus, the way to do it is to pass an object with the values extracted
>> (e.g., a valued sociomatrix or sna edgelist, as noted above). If you have
>> encoded your edge values into a network object, this can easily be done
>> with code like so:
>> library(sna)
>> data(emon)
>> betweenness(as.edgelist.sna(emon[[1]],"Frequency"), ignore.eval=FALSE)
>>
>> In this case, "Frequency" is a numeric variable associated with the edges
>> in emon[[1]]. If we left ignore.eval at its default value of TRUE, we
>> would obtain the corresponding scores for the underlying (unvalued)
>> network. Likewise, if we just called betweenness(emon[[1]]) - with or
>> without ignore.eval=FALSE - the betweenness function would have no idea
>> that we wanted the "Frequency" variable, and would hence operate on the raw
>> graph structure. This same approach is used by most of the functional
>> indices in statnet, but you should always check the individual help pages
>> for details; some indices are not defined for valued data, and others may
>> have additional options that are important to consider.
>>
>> In line with that last note - and mostly for the benefit of others who
>> may encounter the thread - I would also observe that generalizing node or
>> graph-level indices to the valued case is not always trivial, and in some
>> cases the "valued" version of an index may only make sense when the edge
>> values are interpreted in a particular way. For instance, indices based on
>> geodesics (including betweenness) implicitly treat edge values as
>> distances, with the additional implicit assumption that a length of an i,j
>> path is the sum of the values of the edges contained within it. If one's
>> data were to consist e.g. of subjective ratings of positive feeling (e.g.,
>> how much does ego "like" alter on some scale or other), then using these
>> values in this way would make little sense; statnet would dutifully compute
>> the indices for you in such a situation if you asked it to, but they would
>> have no obvious substantive interpretation. In some cases, it may be
>> possible to transform one's original observations into a form that is more
>> appropriate for such an analysis (e.g., transforming similarity scores to
>> distance scores), but in other cases there may be a mismatch between the
>> data properties tacitly assumed by the indices (at least, in their
>> conventional interpretation*) and what was actually measured. In such
>> cases, it may be wise to consider a different index that is more
>> appropriate to one's problem. Finally, it can also be important to ensure
>> that valued data is coded in the way that one intends. For instance, the
>> Drabek et al. EMON data used above codes interaction frequency from 1 to 4,
>> with 1 being the highest level of frequency ("continuous") and four being
>> the lowest non-zero value ("about once a day or less"). For betweenness or
>> closeness (where we are treating the edges as if they are distances), that
>> at least makes ordinal sense, but calculating valued degree without
>> recoding would give us rather perverse results. (Whether it is reasonable
>> to treat path lengths as additive with respect to such a measure is an
>> interesting question that I will not answer, but instead allow to hang in
>> the proverbial air for silent contemplation.) So anyway, be sure to verify
>> that your index is using your edge values in a way that both makes
>> substantive sense and that is compatible with how they are coded.
>>
>> Hope that helps,
>>
>> -Carter
>>
>> * Technically speaking, an index doesn't "assume" anything. It is simply
>> a function of the network. But our conventional *interpretations* of
>> what structural indices tell us about social networks usually involve quite
>> a few assumptions, often tacit. When working with valued data, it is often
>> necessary to become more explicit about what one is assuming.
>> On 6/6/20 7:47 AM, CHU-DING LING wrote:
>>
>> Dear all,
>>
>>
>>
>> I am trying to calculate the betweenness centrality with valued ties, but
>> I do not figure it out. Here are the dataset and codes.
>>
>>
>>
>> 1. This is an example dataset with only four nodes/individuals. When
>> collecting the network data, I asked the participants to answer the
>> question about friendship tie on a 7-point Likert scale. So, this is a
>> *directed* and *valued* network. Moreover, when inputting the network
>> data, I adopted the edge list format and saved it into a CSV file. The
>> details of the data are as follows:
>>
>>
>>
>> Actor Target Friend
>>
>> 1001 1002 5
>>
>> 1001 1003 6
>>
>> 1001 1004 5
>>
>> 1002 1001 6
>>
>> 1002 1003 6
>>
>> 1002 1004 6
>>
>> 1003 1001 4
>>
>> 1003 1002 4
>>
>> 1003 1004 4
>>
>> 1004 1001 6
>>
>> 1004 1002 6
>>
>> 1004 1003 6
>>
>>
>>
>> 2. Then I ran the following codes to calculate the betweenness centrality:
>>
>>
>>
>> library(statnet)
>>
>>
>>
>> #Step 1. read the edgelist format dataset into R
>>
>> Mydata <- read.table("Example.csv", header=TRUE, sep=",")
>>
>>
>>
>> #Step 2. convert an edgelist matrix with valued edges/ties into a network
>>
>>
>>
>> Mynet <- network (Mydata)
>>
>> #Step 3. calculate betweenness centrality but fail to account for the
>> value/weight of the tie
>>
>> betweenness (Mynet)
>>
>>
>>
>> 3. The results came out are as follows:
>>
>>
>>
>> [1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0
>>
>> [43] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [85] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [127] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [169] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [211] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [253] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [295] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [337] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [379] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [421] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [463] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [505] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [547] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [589] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [631] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [673] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [715] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [757] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [799] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [841] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [883] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [925] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0
>>
>> [967] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0
>>
>> [ reached getOption("max.print") -- omitted 4 entries ]
>>
>>
>>
>>
>>
>> It seems that the package treated the IDs of actor and target as edges
>> when computing. *If I want to keep the numeric IDs for the further
>> merging with other variables, what can I do to solve this problem?*
>>
>>
>>
>> Also, if I replace the numeric IDs with strings and organize the data as
>> follows:
>>
>>
>>
>> Actor Target Friend
>>
>> A B 5
>>
>> A C 6
>>
>> A D 5
>>
>> B A 6
>>
>> B C 6
>>
>> B D 6
>>
>> C A 4
>>
>> C B 4
>>
>> C D 4
>>
>> D A 6
>>
>> D B 6
>>
>> D C 6
>>
>>
>>
>> Then, I re-ran the Step 2 and Step 3, the results were as follows:
>>
>>
>>
>> [1] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 44.0 44.0 44.0
>>
>>
>>
>> I think the results are also wrong since the expected results should be
>> four values. However, there are 15 values with the first 12 looking equal.
>>
>>
>>
>> I have searched archival of the list, but I failed to locate the
>> information that can completely solve my problems. So, I am wondering
>> whether any colleagues here could share with me any information about this.
>> I would be grateful if you can provide me any suggestions or references.
>> Many thanks in advance!
>>
>>
>>
>> Best,
>>
>> Chuding
>>
>> _______________________________________________
>> statnet_help mailing liststatnet_help@u.washington.eduhttp://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>>
>> _______________________________________________
>> statnet_help mailing list
>> statnet_help@u.washington.edu
>> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From buttsc at uci.edu Sat Jun 6 20:55:50 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Help seeking on calculating betweenness
centrality with valued ties
In-Reply-To:
References:
<31086dbb-2bea-b400-198c-f178d9094232@uci.edu>
Message-ID:
Hi, Chuding -
I am not sure here to what type of object you are referring - what the
software computes will depend upon what you asked it to do.? If you e.g.
create an adjacency matrix full of zeros, then all functions are going
to treat that as a null graph.? If you create a network object in which
all possible edges are present but then give each edge a numeric
attribute whose value is zero, then (1) functions told to act only on
the raw network structure will treat it as a complete graph, but (2)
extracting the edge values to an adjacency or edgelist matrix will then
lead to the behavior described above.? If one only has one set of edge
values, one usually does not create edges for the zero-valued dyads (in
which case this effect will not occur).? However, one certainly can
create an object like that if one likes (and there are some special
cases in which one may want to do so).
Hope that helps,
-Carter
On 6/6/20 8:27 PM, CHU-DING LING wrote:
>
> Another problem is that even though I transform those valued ties into
> binary ties, the package would still treat everyone in the network is
> connected with the others, as long as I have 12 edges in the 4-vertex
> graph. I tried to remove the cases with a value of edge as 0 in the
> graph and calculated the indegree of each vertex, the results are
> consistent with the reality in the network. So, it seems that the
> package would always treat the column regarding the value of the edges
> as a weight no matter it is a binary or continuous variable. Am I right?
>
> Best,
>
> Chuding
>
>
>
> CHU-DING LING >
> ?2020?6?7??? ??9:44???
>
> Dear Carter,
>
> Thanks for your prompt reply. Actually, I have noticed the
> tutorial on valued ties. However, I did not follow many steps of
> it. And I failed to reproduce the results by just copying and
> running the codes in the examples. There are some errors. That is
> why I have to post this question in the list.
>
> After reading your suggestions, I have attempted to convert the
> network object into a matrix:
>
> Mynet.matrix = as.matrix(Mynet)
>
> class (Mynet.matrix)
>
> [1] "matrix" "array"
>
> And I ran the code to calculate the betweenness centrality again:
>
> betweenness (as.edgelist.sna (Mynet.matrix[[1]],"Friend"),
> ignore.eval=FALSE)
>
> Error in as.edgelist.sna (Mynet.matrix[[1]], "Friend"):
>
> as.edgelist.sna input must be an adjacency matrix/array, edgelist
> matrix, network, or sparse matrix, or list thereof.
>
> According to the results from the ?class (Mynet.matrix)? command,
> I think the ?Mynet.matrix? object is a matrix now. *Why does the
> ?as.edgelist.sna ()? cannot work it out?*
>
> Also, I have explored the class of the ?emon? object in your
> example and results showed it is a list.
>
> class(emon)
>
> [1] "list"
>
> *Does this difference lead the ?as.edgelist.sna ()? command to
> reporting the errors?*
>
> I am sorry to bother you again. I am a novice of R and ?statnet?,
> so I am wondering whether you can give me more detailed
> illustration based on my example dataset. Many thanks in advance!
>
> Best,
>
> Chuding
>
>
>
> Carter T. Butts >
> ?2020?6?7??? ??5:54???
>
> Hi, Chuding -
>
> You may find the valued network tutorial to be helpful here:
> http://statnet.org/Workshops/valued.html
>
> If you are interested in calculating betweenness per se,
> however, it is probably most straightforward to either store
> your data in either a valued adjacency matrix, or an sna
> edgelist (see ?as.edgelist.sna). When you pass the data object
> to the betweenness function, you will also need to tell it to
> use the edge values, since it otherwise will assume that you
> want to calculate betweenness on the underlying unvalued
> digraph (see ?betweenness).
>
> Note that directly passing a network object that happens to
> contain edge values to betweenness et al. will not result in
> those values being used; the reason is that a network object
> can contain any number of different sets of edge variables
> (which need not even be "values" as such - you can have a
> dozen network objects attached to each edge, if you like), and
> the NLI functions have no way of knowing which if any values
> you want it to use.? Thus, the way to do it is to pass an
> object with the values extracted (e.g., a valued sociomatrix
> or sna edgelist, as noted above). If you have encoded your
> edge values into a network object, this can easily be done
> with code like so:
>
> library(sna)
> data(emon)
> betweenness(as.edgelist.sna(emon[[1]],"Frequency"),
> ignore.eval=FALSE)
>
> In this case, "Frequency" is a numeric variable associated
> with the edges in emon[[1]].? If we left ignore.eval at its
> default value of TRUE, we would obtain the corresponding
> scores for the underlying (unvalued) network.? Likewise, if we
> just called betweenness(emon[[1]]) - with or without
> ignore.eval=FALSE - the betweenness function would have no
> idea that we wanted the "Frequency" variable, and would hence
> operate on the raw graph structure. This same approach is used
> by most of the functional indices in statnet, but you should
> always check the individual help pages for details; some
> indices are not defined for valued data, and others may have
> additional options that are important to consider.
>
> In line with that last note - and mostly for the benefit of
> others who may encounter the thread - I would also observe
> that generalizing node or graph-level indices to the valued
> case is not always trivial, and in some cases the "valued"
> version of an index may only make sense when the edge values
> are interpreted in a particular way.? For instance, indices
> based on geodesics (including betweenness) implicitly treat
> edge values as distances, with the additional implicit
> assumption that a length of an i,j path is the sum of the
> values of the edges contained within it.? If one's data were
> to consist e.g. of subjective ratings of positive feeling
> (e.g., how much does ego "like" alter on some scale or other),
> then using these values in this way would make little sense;
> statnet would dutifully compute the indices for you in such a
> situation if you asked it to, but they would have no obvious
> substantive interpretation.? In some cases, it may be possible
> to transform one's original observations into a form that is
> more appropriate for such an analysis (e.g., transforming
> similarity scores to distance scores), but in other cases
> there may be a mismatch between the data properties tacitly
> assumed by the indices (at least, in their conventional
> interpretation*) and what was actually measured.? In such
> cases, it may be wise to consider a different index that is
> more appropriate to one's problem.? Finally, it can also be
> important to ensure that valued data is coded in the way that
> one intends.? For instance, the Drabek et al. EMON data used
> above codes interaction frequency from 1 to 4, with 1 being
> the highest level of frequency ("continuous") and four being
> the lowest non-zero value ("about once a day or less").? For
> betweenness or closeness (where we are treating the edges as
> if they are distances), that at least makes ordinal sense, but
> calculating valued degree without recoding would give us
> rather perverse results.? (Whether it is reasonable to treat
> path lengths as additive with respect to such a measure is an
> interesting question that I will not answer, but instead allow
> to hang in the proverbial air for silent contemplation.)? So
> anyway, be sure to verify that your index is using your edge
> values in a way that both makes substantive sense and that is
> compatible with how they are coded.
>
> Hope that helps,
>
> -Carter
>
> * Technically speaking, an index doesn't "assume" anything.?
> It is simply a function of the network. But our conventional
> /interpretations/ of what structural indices tell us about
> social networks usually involve quite a few assumptions, often
> tacit. When working with valued data, it is often necessary to
> become more explicit about what one is assuming.
>
> On 6/6/20 7:47 AM, CHU-DING LING wrote:
>>
>> Dear all,
>>
>> I am trying to calculate the betweenness centrality with
>> valued ties, but I do not figure it out. Here are the dataset
>> and codes.
>>
>> 1. This is an example dataset with only four
>> nodes/individuals. When collecting the network data, I asked
>> the participants to answer the question about friendship tie
>> on a 7-point Likert scale. So, this is a *directed* and
>> *valued* network. Moreover, when inputting the network data,
>> I adopted the edge list format and saved it into a CSV file.
>> The details of the data are as follows:
>>
>> Actor Target?? Friend
>>
>> 1001 1002? 5
>>
>> 1001 1003? 6
>>
>> 1001 1004? 5
>>
>> 1002 1001? 6
>>
>> 1002 1003? 6
>>
>> 1002 1004? 6
>>
>> 1003 1001? 4
>>
>> 1003 1002? 4
>>
>> 1003 1004? 4
>>
>> 1004 1001? 6
>>
>> 1004 1002? 6
>>
>> 1004 1003? 6
>>
>> 2. Then I ran the following codes to calculate the
>> betweenness centrality:
>>
>> library(statnet)
>>
>> #Step 1. read the edgelist format dataset into R
>>
>> Mydata <- read.table("Example.csv", header=TRUE, sep=",")
>>
>> #Step 2. convert an edgelist matrix with valued edges/ties
>> into a network
>>
>> Mynet <- network (Mydata)
>>
>> #Step 3. calculate betweenness centrality but fail to account
>> for the value/weight of the tie
>>
>> betweenness (Mynet)
>>
>> 3. The results came out are as follows:
>>
>> [1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [43] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [85] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [127] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [169] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [211] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [253] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [295] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [337] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [379] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [421] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [463] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [505] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [547] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [589] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [631] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [673] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [715] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [757] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [799] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [841] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [883] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [925] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>>
>> [967] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> 0 0 0 0 0 0
>>
>> [ reached getOption("max.print") -- omitted 4 entries ]
>>
>> It seems that the package treated the IDs of actor and target
>> as edges when computing. *If I want to keep the numeric IDs
>> for the further merging with other variables, what can I do
>> to solve this problem?*
>>
>> Also, if I replace the numeric IDs with strings and organize
>> the data as follows:
>>
>> Actor Target Friend
>>
>> A? B? 5
>>
>> A? C? 6
>>
>> A? D? 5
>>
>> B? A? 6
>>
>> B? C? 6
>>
>> B? D? 6
>>
>> C? A? 4
>>
>> C? B? 4
>>
>> C? D? 4
>>
>> D? A? 6
>>
>> D? B? 6
>>
>> D? C? 6
>>
>> Then, I re-ran the Step 2 and Step 3, the results were as
>> follows:
>>
>> [1] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 44.0 44.0
>> 44.0
>>
>> I think the results are also wrong since the expected results
>> should be four values. However, there are 15 values with the
>> first 12 looking equal.
>>
>> I have searched archival of the list, but I failed to locate
>> the information that can completely solve my problems. So, I
>> am wondering whether any colleagues here could share with me
>> any information about this. I would be grateful if you can
>> provide me any suggestions or references. Many thanks in advance!
>>
>> Best,
>>
>> Chuding
>>
>>
>> _______________________________________________
>> statnet_help mailing list
>> statnet_help@u.washington.edu
>> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
>
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From malena.haenni at unisg.ch Tue Jun 9 04:34:17 2020
From: malena.haenni at unisg.ch (Haenni, Malena)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Meaning and use of nodesqrtcovar (centered vs.
uncentered)
Message-ID:
Dear statnet-community
I am working with valued ERGMs with count data on public policy cooperation networks of substates. My networks (26 nodes, 325 edges) are undirected and full graphs. To model network dependencies I include sum (obviously), transitiveties (at a given threshold) and nodesqrtcovar to. Here my questions.
What is the difference between the centered and the uncentered term of nodesqrtcovar? When should I use which?
Also, based on the explanations in the help menu, I still struggle to understand the intuition behind nodesqrtcovar. Based on the explanation that it is a valued analog of a 2-star/kstar(2) I understand this as a way to capture activity/popularity of particular nodes. Is this correct? I would very much appreciate an example of a more intuitive way how to interpret this term.
And a big "thank you!" to the developers' team and much appreciation for all the effort you put into it. Ever since the newest release my problems with the traceplots for valued ERGM vanished. I am so grateful!
Kind regards
Malena
Malena Haenni
Phd candidate
[https://owa.unisg.ch/owa/service.svc/s/GetFileAttachment?id=AAMkAGVkMjc4ZDQ2LWE2MTgtNGU2Zi1iMjM0LWMwYzM3N2RkNGM3MABGAAAAAACiVTGePE42TKTx82ovfOpiBwDZ09ADcZZYTK%2F53pderis%2BAAAAAAEJAADZ09ADcZZYTK%2F53pderis%2BAAAAAG1kAAABEgAQAHPgAW%2Fgj3xGso5x3j7BpJI%3D&X-OWA-CANARY=0fTg9RdP3kSUDpQ7yUTodZV9c_1tetQIsS4O3xbzUQzQbfvE3V3TsksfqzFbVgKFFsv07hIw8IM.]
Institut f?r Systemisches Management und Public Governance | (IMP-HSG)
Universit?t St.Gallen (HSG) | Dufourstr. 40a | CH-9000 St.Gallen
Tel. +41 79 796 50 40
malena.haenni@unisg.ch | www.imp.unisg.ch
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From akhanna at medicine.bsd.uchicago.edu Thu Jun 18 18:57:38 2020
From: akhanna at medicine.bsd.uchicago.edu (Khanna, Aditya [MED])
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Relaxing default convergence criteria on ERGMs
Message-ID:
Hello All ?
I am attempting to fit a dyadic dependent ERGM on a network with 32,000 nodes. The model converges once:
The log-likelihood improved by 1.231.
Step length converged once. Increasing MCMC sample size.
But then the job is killed because it runs up against the maximum wall-time on my system (36 hours). I am wondering if there is a simple modification I can make to the control parameters to stop the fitting process after the step length converges the first time? I would like to, if possible, examine the networks that can be simulated from the fit after the first time the step length converges and compare the generated structure to my targets.
The Rout from this session is available at https://gist.github.com/khanna7/11251361f3ef92d2240ddf23f24f1999.
Many thanks for your time,
Aditya Khanna
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From viveksck at gmail.com Mon Jun 29 10:34:11 2020
From: viveksck at gmail.com (Vivek Kulkarni)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] TERGM for weighted networks
Message-ID:
Hi,
I have several snapshots of a weighted network through time and I wanted to
analyze the formation and decay of edges in this temporal network. I
notice that TERGM can help me do that, but I was wondering if TERGm also
supports weighted graphs/networks? I know the package "valued" supports
weighted edges, but then it does not seem to support temporal networks.
Thanks a ton!
Best,
Vivek.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From janbucher at gmail.com Tue Jun 30 00:10:26 2020
From: janbucher at gmail.com (Jan Bucher)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] TERGM for weighted networks
In-Reply-To:
References:
Message-ID:
Hey Vivek,
not an expert here, but when I was faced with the same scenario, I
resorted to flatten the networks using a threshold value. One crucial
question to ask is whether the information in the networks is
dependent or independent within the time slots a single weighted
network would represent.
Best,
Jan
On Mon, Jun 29, 2020 at 7:36 PM Vivek Kulkarni wrote:
>
> Hi,
> I have several snapshots of a weighted network through time and I wanted to analyze the formation and decay of edges in this temporal network. I notice that TERGM can help me do that, but I was wondering if TERGm also supports weighted graphs/networks? I know the package "valued" supports weighted edges, but then it does not seem to support temporal networks. Thanks a ton!
>
> Best,
> Vivek.
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
From david.fisher at abdn.ac.uk Fri Jul 3 09:31:03 2020
From: david.fisher at abdn.ac.uk (Fisher, David)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Estimating difference in event size between 2 levels
of an actor attribute in bipartite networks
Message-ID:
Hi folks,
I hope you are all well. My name is David, I'm a Research fellow at the University of Aberdeen in the UK.
I've recently started playing around with modelling actor by event bipartite networks using "ergm". My first mode are actors with a single attribute, which is a 2-level factor (lets call it "colour", with reds and blues), while my 2nd mode are events which have no attributes. I am interested in two things:
1. Do actors of one category (e.g. reds) appear in more events than actors of the other category?
I am reasonably happy I get at this with "b1factor("colour")"
1. Do actors of one category appear in larger/more well attended events than actors of the other category?
I cannot work out how to model this, but I am certain that there is a term for it and I am just not understanding the documentation. I don't think it is "b2factor("colour")" as events do not vary by colour. I don't think it is any of the nodematch functions (e.g. "b1nodematch("sex")") as I am not interested (for now) in whether reds attend events with other reds. I think I want to compare the number of two paths between the two groups, but I can't see a function to do that. Alternatively, perhaps you could frame it as a b2 k-star question, and does the number of different k-stars where the events are the central node depend on the category of those nodes they are attached too...
Any help would be greatly appreciated.
Best regards,
David
David N. Fisher
Research Fellow
The School of Biological Sciences
University of Aberdeen
Web | GS | Tw | RG | Or
The University of Aberdeen is a charity registered in Scotland, No SC013683.
Tha Oilthigh Obar Dheathain na charthannas cl?raichte ann an Alba, ?ir. SC013683.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From p.krivitsky at unsw.edu.au Sat Jul 4 02:45:05 2020
From: p.krivitsky at unsw.edu.au (Pavel Krivitsky)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] TERGM for weighted networks
In-Reply-To:
References:
Message-ID:
Dear Vivek,
ERGM-based modelling of dynamic weighted network is an open problem,
which I am hoping to tackle in the next year or so. In the meantime,
can you tell me what the weights are in this case? Counts? Categories?
Ranks? Values between 0 and 1?
Best,
Pavel
On Mon, 2020-06-29 at 10:34 -0700, Vivek Kulkarni wrote:
> Hi,
> I have several snapshots of a weighted network through time and I
> wanted to analyze the formation and decay of edges in this
> temporal network. I notice that TERGM can help me do that, but I was
> wondering if TERGm also supports weighted graphs/networks? I know the
> package "valued" supports weighted edges, but then it does not seem
> to support temporal networks. Thanks a ton!
>
> Best,
> Vivek.
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
From viveksck at gmail.com Sat Jul 4 08:17:20 2020
From: viveksck at gmail.com (Vivek Kulkarni)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] TERGM for weighted networks
In-Reply-To:
References:
Message-ID:
Hi Pavel,
Thanks for letting me know! In my setting, the weights are just counts that
represent the number of observed co-occurrences between words in a
co-occurrence network.
Best,
Vivek.
On Sat, Jul 4, 2020 at 2:45 AM Pavel Krivitsky
wrote:
> Dear Vivek,
>
> ERGM-based modelling of dynamic weighted network is an open problem,
> which I am hoping to tackle in the next year or so. In the meantime,
> can you tell me what the weights are in this case? Counts? Categories?
> Ranks? Values between 0 and 1?
>
> Best,
> Pavel
>
> On Mon, 2020-06-29 at 10:34 -0700, Vivek Kulkarni wrote:
> > Hi,
> > I have several snapshots of a weighted network through time and I
> > wanted to analyze the formation and decay of edges in this
> > temporal network. I notice that TERGM can help me do that, but I was
> > wondering if TERGm also supports weighted graphs/networks? I know the
> > package "valued" supports weighted edges, but then it does not seem
> > to support temporal networks. Thanks a ton!
> >
> > Best,
> > Vivek.
> > _______________________________________________
> > statnet_help mailing list
> > statnet_help@u.washington.edu
> > http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From david.fisher at abdn.ac.uk Mon Jul 6 00:59:06 2020
From: david.fisher at abdn.ac.uk (Fisher, David)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] TERGM for weighted networks
In-Reply-To:
References:
Message-ID:
Hi Vivek (and list),
Stochastic actor oriented models can handle 2 different strengths of ties (last time I checked) with a dynamic framework. Although not ERGMs, it still might be of interest.
Check out the RSiena scripts on this page, http://www.stats.ox.ac.uk/~snijders/siena/ I think script #11 fits a dynamic network model that is ?ordered? (i.e. ties with different weights). You?ll probably need to look at the manual and the more simple scripts first though to make sense of it.
Cheers,
David
From: Vivek Kulkarni
Sent: 04 July 2020 16:17
To: Pavel Krivitsky
Cc: statnet_help@u.washington.edu
Subject: Re: [statnet_help] TERGM for weighted networks
Hi Pavel,
Thanks for letting me know! In my setting, the weights are just counts that represent the number of observed co-occurrences between words in a co-occurrence network.
Best,
Vivek.
On Sat, Jul 4, 2020 at 2:45 AM Pavel Krivitsky > wrote:
Dear Vivek,
ERGM-based modelling of dynamic weighted network is an open problem,
which I am hoping to tackle in the next year or so. In the meantime,
can you tell me what the weights are in this case? Counts? Categories?
Ranks? Values between 0 and 1?
Best,
Pavel
On Mon, 2020-06-29 at 10:34 -0700, Vivek Kulkarni wrote:
> Hi,
> I have several snapshots of a weighted network through time and I
> wanted to analyze the formation and decay of edges in this
> temporal network. I notice that TERGM can help me do that, but I was
> wondering if TERGm also supports weighted graphs/networks? I know the
> package "valued" supports weighted edges, but then it does not seem
> to support temporal networks. Thanks a ton!
>
> Best,
> Vivek.
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
The University of Aberdeen is a charity registered in Scotland, No SC013683.
Tha Oilthigh Obar Dheathain na charthannas cl?raichte ann an Alba, ?ir. SC013683.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From adamhaber at gmail.com Wed Jul 8 13:15:03 2020
From: adamhaber at gmail.com (Adam Haber)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Distance-dependent ERGM model
Message-ID: <8BBB17DE-C891-4FFF-B5BB-8B7D0ABA1710@gmail.com>
Hello,
(Apologies in advance if this is a trivial question - I?m quite new to modelling with ERGMs, and I couldn?t find an answer elsewhere...)
The nodes in the network I?m studying are embedded in 3D, and I have reasons to believe that the spatial organisation plays a major role in the structure of the network, along with a specific node-level attribute. How can one model such a spatial dependency using ergm? AFAIU, it?s not absdiff, since it?a a function of 3 numeric node-level attributes (it?s x-y-z coordinates). Is it also possible to have a different distance-dependence for different levels of some node-level attribute?
Any help would be much appreciated!
Best regards,
Adam Haber
Department of Neurobiology
Weizmann Institute of Science
From buttsc at uci.edu Wed Jul 8 18:31:00 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Distance-dependent ERGM model
In-Reply-To: <8BBB17DE-C891-4FFF-B5BB-8B7D0ABA1710@gmail.com>
References: <8BBB17DE-C891-4FFF-B5BB-8B7D0ABA1710@gmail.com>
Message-ID: <42c1087a-0424-2f87-5ad7-bb345425c144@uci.edu>
Hi, Adam -
Distance-based models are indeed a thing.? Although there many are other
things you can do, the easiest starting point is usually to calculate
the distances among all vertices in whatever metric is most apposite (in
your case, Euclidean space), and use the distance matrix or some
function thereof as an edgecov.? Using the raw distances as an edge
covariate implicitly models a logistic effect of distances on tie
probability; it can make sense when ties are necessarily local.? In
social settings, it usually makes more sense to work with the log
distances, which approximates a power-law decline in marginal tie
probability as a function of distance.? If you can effectively draw a
sphere around a node and say to yourself (and others), "well, the chance
of there being any ties to nodes outside this sphere is essentially
zero" (i.e., probability small and falling exponentially), then the draw
model is a reasonable approximation.? If not (i.e., there's always some
non-vanishing chance of a long range tie, even if the probability is
very low), then you want the log version.? (Yes, this is heuristic, and
there are other things that one can do.? But I would start with this.)
Hope that helps,
-Carter
On 7/8/20 1:15 PM, Adam Haber wrote:
> Hello,
>
> (Apologies in advance if this is a trivial question - I?m quite new to modelling with ERGMs, and I couldn?t find an answer elsewhere...)
>
> The nodes in the network I?m studying are embedded in 3D, and I have reasons to believe that the spatial organisation plays a major role in the structure of the network, along with a specific node-level attribute. How can one model such a spatial dependency using ergm? AFAIU, it?s not absdiff, since it?a a function of 3 numeric node-level attributes (it?s x-y-z coordinates). Is it also possible to have a different distance-dependence for different levels of some node-level attribute?
>
> Any help would be much appreciated!
>
> Best regards,
> Adam Haber
> Department of Neurobiology
> Weizmann Institute of Science
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
From david.fisher at abdn.ac.uk Mon Jul 13 05:16:57 2020
From: david.fisher at abdn.ac.uk (Fisher, David)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Estimating difference in event size between 2
levels of an actor attribute in bipartite networks
In-Reply-To:
References:
Message-ID:
Any ideas? If my question is unclear then perhaps I can re-phrase it?
Cheers,
David
From: Fisher, David
Sent: 03 July 2020 17:31
To: statnet_help@u.washington.edu
Subject: [statnet_help] Estimating difference in event size between 2 levels of an actor attribute in bipartite networks
Hi folks,
I hope you are all well. My name is David, I'm a Research fellow at the University of Aberdeen in the UK.
I've recently started playing around with modelling actor by event bipartite networks using "ergm". My first mode are actors with a single attribute, which is a 2-level factor (lets call it "colour", with reds and blues), while my 2nd mode are events which have no attributes. I am interested in two things:
1. Do actors of one category (e.g. reds) appear in more events than actors of the other category?
I am reasonably happy I get at this with "b1factor("colour")"
1. Do actors of one category appear in larger/more well attended events than actors of the other category?
I cannot work out how to model this, but I am certain that there is a term for it and I am just not understanding the documentation. I don't think it is "b2factor("colour")" as events do not vary by colour. I don't think it is any of the nodematch functions (e.g. "b1nodematch("sex")") as I am not interested (for now) in whether reds attend events with other reds. I think I want to compare the number of two paths between the two groups, but I can't see a function to do that. Alternatively, perhaps you could frame it as a b2 k-star question, and does the number of different k-stars where the events are the central node depend on the category of those nodes they are attached too...
Any help would be greatly appreciated.
Best regards,
David
David N. Fisher
Research Fellow
The School of Biological Sciences
University of Aberdeen
Web | GS | Tw | RG | Or
The University of Aberdeen is a charity registered in Scotland, No SC013683.
Tha Oilthigh Obar Dheathain na charthannas cl?raichte ann an Alba, ?ir. SC013683.
The University of Aberdeen is a charity registered in Scotland, No SC013683.
Tha Oilthigh Obar Dheathain na charthannas cl?raichte ann an Alba, ?ir. SC013683.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From dluke at wustl.edu Mon Jul 13 12:13:50 2020
From: dluke at wustl.edu (Luke, Douglas)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Estimating difference in event size between 2
levels of an actor attribute in bipartite networks
In-Reply-To:
References: ,
Message-ID:
David,
If I'm following this correctly, then you will want the event mode to have an attribute as well. So, if you want to know if reds/blues are attending larger events in greater numbers, then the event mode will have to have some type of 'size' attribute. That should open up some analytic possibilities for you.
--Doug--
Douglas Luke
Director, Center for Public Health Systems Science
Professor of Public Health, Brown School
Washington University in St. Louis
Campus Box 1196, One Brookings Drive
St. Louis, MO 63130
email: dluke@wustl.edu
Website: http://cphss.wustl.edu
________________________________
From: Fisher, David
Sent: Monday, July 13, 2020 7:16 AM
To: statnet_help@u.washington.edu
Subject: Re: [statnet_help] Estimating difference in event size between 2 levels of an actor attribute in bipartite networks
Any ideas? If my question is unclear then perhaps I can re-phrase it?
Cheers,
David
From: Fisher, David
Sent: 03 July 2020 17:31
To: statnet_help@u.washington.edu
Subject: [statnet_help] Estimating difference in event size between 2 levels of an actor attribute in bipartite networks
Hi folks,
I hope you are all well. My name is David, I?m a Research fellow at the University of Aberdeen in the UK.
I?ve recently started playing around with modelling actor by event bipartite networks using ?ergm?. My first mode are actors with a single attribute, which is a 2-level factor (lets call it ?colour?, with reds and blues), while my 2nd mode are events which have no attributes. I am interested in two things:
1. Do actors of one category (e.g. reds) appear in more events than actors of the other category?
I am reasonably happy I get at this with ?b1factor("colour")?
1. Do actors of one category appear in larger/more well attended events than actors of the other category?
I cannot work out how to model this, but I am certain that there is a term for it and I am just not understanding the documentation. I don?t think it is ?b2factor(?colour?)? as events do not vary by colour. I don?t think it is any of the nodematch functions (e.g. ?b1nodematch("sex")?) as I am not interested (for now) in whether reds attend events with other reds. I think I want to compare the number of two paths between the two groups, but I can?t see a function to do that. Alternatively, perhaps you could frame it as a b2 k-star question, and does the number of different k-stars where the events are the central node depend on the category of those nodes they are attached too?
Any help would be greatly appreciated.
Best regards,
David
David N. Fisher
Research Fellow
The School of Biological Sciences
University of Aberdeen
Web | GS | Tw | RG | Or
The University of Aberdeen is a charity registered in Scotland, No SC013683.
Tha Oilthigh Obar Dheathain na charthannas cl?raichte ann an Alba, ?ir. SC013683.
The University of Aberdeen is a charity registered in Scotland, No SC013683.
Tha Oilthigh Obar Dheathain na charthannas cl?raichte ann an Alba, ?ir. SC013683.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From adamhaber at gmail.com Mon Jul 13 12:48:09 2020
From: adamhaber at gmail.com (Adam Haber)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Hyperparameters for latent position models
Message-ID: <5FD9E534-AB07-4291-B6ED-CBBBD0EA9288@gmail.com>
Hello,
I?m using ergmm to fit latent position models. So far I?ve chosen d (embedding dimension) and G (number of clusters) based on domain-knowledge consideration, and inspection of the GOF plots/stats of different (d,G) pairs. Is there a more ?principled? approach to choosing these hyperparameters? Is there anything similar to cross-validation for choosing hyper parameters in this class of models?
Any help would be much appreciated!
Best regards,
Adam Haber
Department of Neurobiology
Weizmann Institute of Science
From adamhaber at gmail.com Sun Jul 19 03:07:23 2020
From: adamhaber at gmail.com (Adam Haber)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Interactions in a distance-dependent model
Message-ID:
Hello,
Following a recent question I?ve posted here (and got very helpful responses - thank you!), I?m trying to add more ?domain knowledge? into the model. Specifically, the nodes in the network we're studying are embedded in 3D, and we?ve seen that (as expected) adding distance-dependence to the model (via an edgecov(x) term such that x is the distance matrix) indeed improves the GOF.
I want to take this one step further: I know that if there?s a (directed) edge i->j, and j and k are ?close? (spatially), it should increase the probability that there?s an edge i->k. Another option would be to ?discretize? the distances and group nodes into groups of ?spatial loci", and add a term that if there?s an edge i->j and j and k are in the same spatial cluster (a binary indicator function), than this should increase the probability of the edge i->k.
Is there a way to incorporate this sort of reasoning into an ergm model using any of the available terms? I went over the examples I could find and did not encounter anything similar...
Thanks again,
Respectfully,
Adam Haber
From buttsc at uci.edu Sun Jul 19 04:15:20 2020
From: buttsc at uci.edu (Carter T. Butts)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] Interactions in a distance-dependent model
In-Reply-To:
References:
Message-ID:
Hi, Adam -
If you have a mutuality term in the model, it already accounts for that
effect.? Do you intend to say that you expect proximate vertices to
reciprocate at higher rates - /above and beyond the propinquity effect/
- than distant vertices?? (Again, if you have a propinquity effect, it
will already be the case that, ceteris paribus, reciprocation will be
more likely for proximate vertices.)
If you want this type of effect, it can be realized with the dyadcov
term, or by creating an interaction term between mutual and edgecov.? I
don't think the user-level functionality for the latter option is yet
exposed, though it's pretty easy to code it using ergm.userterms;
however, dyadcov will probably suit your purposes.
Hope that helps,
-Carter
On 7/19/20 3:07 AM, Adam Haber wrote:
> Hello,
>
> Following a recent question I?ve posted here (and got very helpful responses - thank you!), I?m trying to add more ?domain knowledge? into the model. Specifically, the nodes in the network we're studying are embedded in 3D, and we?ve seen that (as expected) adding distance-dependence to the model (via an edgecov(x) term such that x is the distance matrix) indeed improves the GOF.
>
> I want to take this one step further: I know that if there?s a (directed) edge i->j, and j and k are ?close? (spatially), it should increase the probability that there?s an edge i->k. Another option would be to ?discretize? the distances and group nodes into groups of ?spatial loci", and add a term that if there?s an edge i->j and j and k are in the same spatial cluster (a binary indicator function), than this should increase the probability of the edge i->k.
>
> Is there a way to incorporate this sort of reasoning into an ergm model using any of the available terms? I went over the examples I could find and did not encounter anything similar...
>
> Thanks again,
> Respectfully,
> Adam Haber
>
> _______________________________________________
> statnet_help mailing list
> statnet_help@u.washington.edu
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From adamhaber at gmail.com Sun Jul 19 12:43:48 2020
From: adamhaber at gmail.com (Adam Haber)
Date: Tue Aug 3 21:58:15 2021
Subject: [statnet_help] statnet_help Digest, Vol 167, Issue 7
In-Reply-To:
References:
Message-ID: <05FE7BDC-FF03-4857-AADC-E33D5D68BE0E@gmail.com>
Hi Carter,
Thanks (again!) for the detailed response. I?m probably missing something, but I?m not sure I understand why mutuality would be the relevant term here. Let?s say that my nodes are embedded in 2D, and node i is at the origin. I also have two other nodes - node j at (1,0) and node k at (1.001, 0) - such that j and k are very close, relative to i. I?m trying to build a model in which the edges (i->j) and (i->k) are dependent - if one of them exists, the other is more likely to exists as well. Mathematically, I think a good way to denote what I?m trying to achieve is P(i->k | i->j, j and k are very close) > P(i->k), where P(i->k) is the marginal probability that an edge from i to k exists (already taking into account their distance and potentially the effects of other terms).
If I understand correctly, mutuality would affect P(j->i | i->j), or P(i->k | k-> i), but not P(i->k | i->j). But again, maybe I?m missing something - I?m quite new to ERGMs.
Regarding interaction - I thought one possible way to do this is to ?discretise? the close/far into categories (clusters of nodes), and then add an interaction term between this new ?cluster factor? and relevant triad-related terms, as these seem to capture this dependence I?m after. Like you said, I?m not sure ERGM exposes such functionality.
Thanks again!
Best,
Adam
> On 19 Jul 2020, at 22:01, statnet_help-request@mailman13.u.washington.edu wrote:
>
> Send statnet_help mailing list submissions to
> statnet_help@u.washington.edu
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://mailman13.u.washington.edu/mailman/listinfo/statnet_help
> or, via email, send a message with subject or body 'help' to
> statnet_help-request@mailman13.u.washington.edu
>
> You can reach the person managing the list at
> statnet_help-owner@mailman13.u.washington.edu
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of statnet_help digest..."
> Today's Topics:
>
> 1. Interactions in a distance-dependent model (Adam Haber)
> 2. Re: Interactions in a distance-dependent model (Carter T. Butts)
>
> From: Adam Haber
> Subject: [statnet_help] Interactions in a distance-dependent model
> Date: 19 July 2020 at 13:07:23 GMT+3
> To: statnet_help@u.washington.edu
>
>
> Hello,
>
> Following a recent question I?ve posted here (and got very helpful responses - thank you!), I?m trying to add more ?domain knowledge? into the model. Specifically, the nodes in the network we're studying are embedded in 3D, and we?ve seen that (as expected) adding distance-dependence to the model (via an edgecov(x) term such that x is the distance matrix) indeed improves the GOF.
>
> I want to take this one step further: I know that if there?s a (directed) edge i->j, and j and k are ?close? (spatially), it should increase the probability that there?s an edge i->k. Another option would be to ?discretize? the distances and group nodes into groups of ?spatial loci", and add a term that if there?s an edge i->j and j and k are in the same spatial cluster (a binary indicator function), than this should increase the probability of the edge i->k.
>
> Is there a way to incorporate this sort of reasoning into an ergm model using any of the available terms? I went over the examples I could find and did not encounter anything similar...
>
> Thanks again,
> Respectfully,
> Adam Haber
>
>
>
>
>
> From: "Carter T. Butts"