Statalist


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

st: gllamm for 3 level model with 2 levels of clustering


From   Elizabeth Aura McClintock <emcclint@stanford.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: gllamm for 3 level model with 2 levels of clustering
Date   Fri, 10 Jul 2009 12:48:45 -0700 (PDT)

Hi~
 
I would appreciate any suggestions about how I might estimate a three-level model that adjusts for clustering at two levels (relationship-within-individual and individual-within-school).  Can this be done using GLLAMM (or any other way)?
  
I am trying to estimate a regression model using relationship-level data that I created from individual-level data.  When analyzing individual-level data, the correct survey command for the waves of data that I am using is:
 
svyset [pweight=pw1], strata(st) psu(psu) 
 
But I am analyzing relationship-level data in which a respondent might contribute several observations (if s/he reported several relationships) or no observations (if s/he reported no relationships), so I also need to cluster by respondent.
 
I have tried the following command in GLLAMM:
 
gllamm ...., i(id psu st) pweight(pw) link(logit) family(bin)
 
Where id = respondent ID 
 
Is this correctly specified?  And is there any way to increase the speed of estimation?  I have 3 dependent variables and for one of these outcomes the estimation process seems to be iterating interminably (over a week so far...).  I tried adding the option "nip(4)" but this did not seem to help much.  The sample size is about 20,000.  
 
Thank you in advance for any advice you can give me!
 
~Elizabeth
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



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