st: multilevel multinomial logit predicted probabilities

["Stoeckel, Florian" <fs@unc.edu>]

Dear Statalist,

I'm new to the list, so I apologize if my question has already come up! My
problem concerns the calculation of a confidence interval around the
changes in predicted probabilities in a multilevel multinomial logistic
regression. I use gllamm and know about the gllapred post estimation
command. Unfortunately, I could neither figure out how to calculate the
changes in the (conditional) predicted probabilities with gllapred, nor
how this command would provide me with standard errors to get to the
confidence intervals.

My model is a random intercept model (no random slopes, no cross level
interactions) with four unordered response categories. I want to calculate
the change in the predicted probabilities of the four response categories
when the value of an independent variable changes from one standard
deviation below the mean to one above the mean (holding all other
variables at their mean). To be more precise: I want to show the
percentage changes in the predicted probabilities, rather than the
predicted probabilities.  In order to see if the percent of change in the
predicted probabilities is significant, I want to calculate the 95%
confidence interval around these changes. So far, I can run the model with
gllamm. And I calculated the predicted probabilities and the changes in
Excel, which does however not help me with the standard errors of course.

I would be very grateful for any advice!

Many thanks,

Florian Stoeckel


--
PhD Student 
UNC Chapel Hill
Department of Political Science



*
*   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/