Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

st: Multilevel (2 Level) Multinomial Logit Model


From   Anthony Fulginiti <[email protected]>
To   [email protected]
Subject   st: Multilevel (2 Level) Multinomial Logit Model
Date   Thu, 22 Aug 2013 12:40:02 -0700

Hello Statalist,

I am running a 2 level model for a social network analysis.  Types of relationship ties are at level 1, which are clustered in individuals (Individual Personal Networks to be specific) at Level 2.  My outcome or dependent variables is type of social support (tie) , which has greater than 2 categories (0=no support, 1=advice support, 2=tangible support).  To account for clustering, I want to fit a 2-level random intercept model.  In a recent article by delaHaye et al. (2012), "Who is supporting homeless youth? Predictors of support in personal networks" Journal of Research on Adolescence, 22(4), they employ a multilevel multinomial model using gllamm and I was seeking to follow a similar approach.  

I have been reviewing the gllamm manual (Rabe-Hesketh et al., 2004) as well as Grilli and Rampichini (2005), "A review of random effects modelling using gllamm in Stata", which provide some examples .  I have not had a tremendous amount of experience working with multinomial models so my apologies in advance if I have not yet digested the information offered in these aforesaid references.  

However, on p. 13-14 of the Grilli and Rampichini reference, they provide the following code for a "2-level random effects multinomial model" without covariates (with my variables substituted here):

gllamm supporttype, i(id) base (0) link (mlogit) family(binomial) nip(8) adapt

Question 1:  If I wanted to look at the relationship between being homeless and my social support outcome in the context of a random intercept multinomial model, would the following code be appropriate?  While I have fit multilevel models in the past, this is my first attempt at a multilevel multinomial model.

gllamm supporttype homeless, i(id) base (0) link (mlogit) family(binomial) nip(8) adapt

Question 2: If I wanted to then perform a random coefficient model (to compare with fit with the random intercept model), how could I then run a model suggesting that the effect of homeless (as a level 1 variable) would vary across network clusters?


Thanks in advance for any feedback on either question that you can offer.


Anthony Fulginiti






*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index