Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Some additional questions regarding -gllamm, link (mlogit)-


From   Andreas Schiffelholz <Andreas.Schiffelholz@gmx.de>
To   statalist@hsphsun2.harvard.edu
Subject   st: Some additional questions regarding -gllamm, link (mlogit)-
Date   Fri, 26 Apr 2013 14:05:06 +0200

Hi everyone,

Thanks a lot for the support I received so far, trying to understand the -gllamm- command in all its detail. I have three additional questions regarding options of the -gllamm- command when using it to calculate an multinomial logistic regression with random effects. I already consulted the manual as well as the book of Rabe-Hesketh and Skrondal (Volume II), but did not the options in all detail.

I'm working with a unbalanced panel dataset (x: companies; t:years) and two models, one with two, one with three (nominal) responses (basically one of the two responses of the first model is split into two different responses -> 3 responses). I calculate the ordinal models using a -logit, cluster (company)- approach for pooled and an -xtlogit, re vce(bootstrap)- approach for the random effects model. I want to calculate the nominal model in a comparable way. For the pooled model I use the -mlogit, cluster (company)- command. For the multinomial logistic regression with random effects I want to use a -gllamm- based model:

-gllamm varlist, i(Company) base(0) link(mlogit) cluster(Company) family(binom) nip(12) adapt trace-

In the documentation of gllamm several additional options are used. I do not fully understand if they are needed in my case (my goal is to stay as close and comparable to my -xtlogit, re vce (bootstrap)- approach as possible):

1. -expanded- is this option always needed for mlogit or only if there are different sets of alternatives? (In my dataset, the three alternatives are the same for every company in every year)

2. -nrf (...)- As far as I understand it the default is nrf(1), do I need more than one random effect?

3. -eqs (...)- To be honest, I do not fully understand this option, is it needed in my case?


Andreas
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index