st: Error message resulting from gllapred

["Eberth, Jan Marie" <Jan.M.Eberth@uth.tmc.edu>]

Using STATA, I ran a 2 level model using gllamm (xi: gllamm outcome i.race age, i(cluster) l(logit) f(binom) pweight(pwt) nip(8) adapt) on the estimation sample (i.e. dataset containing 574 persons without any missing data).

After running this model, I opened up the prediction dataset (i.e. dataset with ~7000 persons; race and age present; outcome missing) and tried to run gllapred (options such as mu and linpred). When running gllapred, I recieved the following error message: log-likelihood cannot be computed for observations that should not contribute to posterior, set response variable to missing r(198). Interestingly, the gllapred command seemed to work fine when the gllamm model didn't have the pweight option specified.  It is also noteworthy that I didn't get this error message when running gllasim.

I'd really like to get gllapred working for the prediction dataset using the parameters from the weighted, 2 level model. Does anyone have insight into why I'm getting this error and how I might be able to fix it?

Thank you!
Jan Eberth, MSPH


*
*   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/