[Rsiena-help] Help interpreting the estimates

[Tom Snijders]

Dear Nathan,

This kind of question is more appropriate (and will get more replies) for the list
http://groups.yahoo.com/groups/stocnet/
Also see
http://www.stats.ox.ac.uk/~snijders/siena/siena_main.htm

For the rest: I now have no time to give an answer, and may return to this in two weeks (when I’ll be back from holidays). But at the other list there may well be members who could reply to this.

Best,
Tom


================================================================
Tom A.B. Snijders
Professor of Statistics and Methodology
Dept. of Sociology, University of Groningen
Emeritus Fellow, Nuffield College, University of Oxford
Associate Member, Dept. of Statistics, University of Oxford
http://www.stats.ox.ac.uk/~snijders/

From: rsiena-help-bounces at lists.r-forge.r-project.org [mailto:rsiena-help-bounces at lists.r-forge.r-project.org] On Behalf Of Nathan Abe
Sent: 06 August 2017 05:13
To: rsiena-help at lists.r-forge.r-project.org
Subject: [Rsiena-help] Help interpreting the estimates



down votefavorite<https://stats.stackexchange.com/questions/296424/how-to-interpret-saom-estimates>


I have been reviewing many articles and manuals and have a question regarding the interpretations of SAOM and ERGM-based models. From what I can tell, it seems the SAOM values, say just at the most basic edge level, seem to be systematically different from the edge values in the ERGM-based models in that the SAOM estimates are not as extreme as say TERGM or STERGM estimates. I am observing this in my own research and see the same pattern in Desmarais & Cranmer (2012) article (Micro-Level Interpretation of Exponential Random Graph Models with Application to Estuary Networks). My understanding is that both parameters are reported in log-odds. My intuition is that the interpretation will be different due to the actor-orientation of the SAOM and considering ties from the perspective of each actor, but I can't quite wrap my mind around what the exact interpretation would be in such a way that it explains the systematic difference in values.

Below is an example of some R code and output for the SAOM model.

seed1 <- sample(1:10000, 1, replace = T)



g0 <- network(20, density = 0.36, directed = TRUE)



g1 <- simulate(~edges + mutual, nsim = 1,

                            coef = c(0.1025, 0),

                            basis = g0,

                            control = control.simulate(MCMC.burnin = 1000,

                                      MCMC.interval = 100), seed = seed1)

SAOM

betterAlgorithm <- sienaAlgorithmCreate(projname = "D", diagonalize = 0.2,

                                n3 = 4000)



X1 <- as.matrix(g0)

X2 <- as.matrix(g1)



mynet1 <- sienaDependent(array(c(X1, X2), dim = c(20, 20, 2)))

mydata <- sienaDataCreate(mynet1)

myeff <- getEffects(mydata)

myeff <- includeEffects(myeff, recip)

ans3 <- siena07(betterAlgorithm, data = mydata, effects = myeff, batch = T)

summary(ans3)

## Estimates, standard errors and convergence t-ratios

##

##                               Estimate   Standard   Convergence

##                                            Error      t-ratio

##

## Rate parameters:

##   0       Rate parameter      18.6356  ( 2.8187   )

##   1. eval outdegree (density)  0.0640  ( 0.1067   )   0.0040

##   2. eval reciprocity          0.2621  ( 0.1626   )   0.0231

##

## Total of 5011 iteration steps.

##

## Overall maximum convergence ratio:    0.0369

##

## Covariance matrix of estimates (correlations below diagonal)

##

##        0.011       -0.013

##       -0.753        0.026

##

## Derivative matrix of expected statistics X by parameters:

##

##      137.587      146.741

##       69.844      115.650

##

## Covariance matrix of X (correlations below diagonal):

##

##       93.473      101.707

##        0.844      155.425

I would greatly appreciate any insight on this that you have!
Best,
Nate


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rsiena-help/attachments/20170806/b5189532/attachment-0001.html>