[Rsiena-help] Some parameters are lack of convergence t-ratio

[Shupin Li]

Thank you Tom! 
I went through the scripts of s_50 data (siena example dataset for tutorial) and it used the similar parameters in sienaAlgorithmCreate function (it used drinking and smoking and I used gender as attribute effects, other network effects were the same in the model). I am not sure why outcome of s_50 data has convergence t-ratio for parameters of network dynamics using default setting.
Further, I want to investigate the network dynamic parameters in the model. How shall I set the cond parameter in sienaAlgorithmCreate function in this case?

Best,
Shupin

________________________________________
发件人: Tom Snijders [Tom.Snijders at nuffield.ox.ac.uk]
发送时间: 2015年10月14日 10:46
收件人: Shupin Li; rsiena-help at lists.r-forge.r-project.org
主题: RE: Some parameters are lack of convergence t-ratio

Dear Shupin,

For this type of question, the Stocnet/Siena discussion list is more appropriate:
http://groups.yahoo.com/groups/stocnet/

But let me answer your question here.
The estimation here is done by default by the conditional estimation option. See section 6.9.1 in the manual. For these parameters, there is no approximation involved in the estimation. Therefore convergence indicators are not an issue of concern.

Best wishes,
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/


-----Original Message-----
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 Shupin Li
Sent: 14 October 2015 08:48
To: rsiena-help at lists.r-forge.r-project.org
Subject: [Rsiena-help] Some parameters are lack of convergence t-ratio

Dear folks,

I have data of a 20 people network in 5 time waves and their gender. I wanted to specify models for dynamics of networks with some basic network effects and gender homophily. I used below commands, but some parameters (rate parameters) are lack of convergence t-ratio in the outcome (though the overall maximum convergence ratio looked nice).

Why don't I have convergence t-ratios for those rate parameters? How shall I solve this problem?

> myCoEvolutionEff <- includeEffects( myCoEvolutionEff, transTrip,
> cycle3) myCoEvolutionEff <- includeEffects( myCoEvolutionEff, simX,
> interaction1 = "gender" ) myCoEvAlgorithm <- sienaAlgorithmCreate(
> projname = 'CoEv_5' ) # define the algorithm settings
> betterCoEvAlgorithm <- sienaAlgorithmCreate( projname = 'CoEv_5',
> diagonalize = 0.2, doubleAveraging = 0) (ans <- siena07(
> betterCoEvAlgorithm, data = myCoEvolutionData, effects =
> myCoEvolutionEff )) ans
Estimates, standard errors and convergence t-ratios

                                                  Estimate   Standard        Convergence
                                                                 Error                   t-ratio

Rate parameters:
  0.1      Rate parameter period 1  6.1412  ( 1.2194   )
  0.2      Rate parameter period 2  5.4820  ( 1.1153   )
  0.3      Rate parameter period 3  4.2816  ( 0.8480   )
  0.4      Rate parameter period 4  3.2814  ( 0.6573   )

Other parameters:
  1.  eval outdegree (density)       -1.0531  ( 0.2341   )           -0.0327
  2.  eval reciprocity                      0.7823  ( 0.1933   )           -0.0414
  3.  eval transitive triplets              0.1559  ( 0.0223   )           -0.0400
  4.  eval 3-cycles                         -0.1406  ( 0.0411   )           -0.0349
  5.  eval gender similarity              0.9369  ( 0.1561   )           -0.0906

Overall maximum convergence ratio:    0.1234
Total of 2249 iteration steps.

Any hints are appreciated highly. Thanks a lot in advance!

Best,
Shupin

_______________________________________________
Rsiena-help mailing list
Rsiena-help at lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rsiena-help