st: handling zero random effects automatically in xtmixed

[David Airey <david.airey@vanderbilt.edu>]

.

This is related to my previous post on xtmixed syntax. In a microarray  
data set of gene expression, we are exploring parametric mixed model  
gene set enrichment analysis (GSEA). SAS Proc Mixed seems to bully its  
way through models with random effects that are essentially zero.  
Stata doesn't converge. The solution is to troubleshoot the model,  
maybe using emonly and 500 iterations, to verify a zero random effect,  
remove it, reformulate the model and run it again. The problem is that  
we are talking hundreds if not thousands of variance components mixed  
models. There needs to be an automatic method of dealing with zero  
estimated random effects. I don't know yet what SAS is doing  
differently, but I will find out.

-Dave
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/