st: increasing variance when adding covariates (xtmelogit)

["de Vries, Robert" <r.de-vries08@imperial.ac.uk>]

Hello all,

I am getting some strange results from some multilevel logistic regression models 
(persons nested within geographical areas, random intercepts) run with 
xtmlogit.

I am using xtmelogit to fit models predicting a binary outcome at the 
person level from some person-level and area level predictors. In the 
empty model (with no predictors) I get a certain estimate for the 
standard deviation of the constant. However, when adding
person and country-level predictors to the model, the estimate of this 
property actually INCREASES quite substantially (by about 15-20%).

As I take the standard deviation of the constant to be an indicator of 
(the square root of) the variance at the area level. It seems very 
strange that this would go up when adding predictors to the model. I 
would be grateful if anyone could shed any light on
this situation.

Rob
*
*   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/