st: xtlogit re versus xtlogit pa

["Christer Thrane" <ch-thran@online.no>]

Hi,

My observations are person/years and the dependent variable is whether a 
person in any given year is registrated as "in continuing education, 
full-time" (= 1) versus "not" (= 0); that is, still is in regular 
employment. The independent variables are gender, work experince in years, 
time (i.e. observation year plus some controls.

My "problem" is that the predicted probabilities from respectively -re- 
and -pa- differ substantially.

1. Why does the -re- model, for various combinations of the x-values, in 
general produce lower predicted probabilities than the -pa- model? (By the 
way, the results for the -pa- model is more in agreement with the 
plain -logit, cluster (id)- model's results than the -re- model.)

2. Only the -pa- model has a robust option. But why would do I need a robust 
option in this case? Does not the -xt- option "solve" the issue of many 
observations for the same person. Or does the robust option solve some other 
issue in this regard, such as arbitrary serial correlation between panels?

Thanks in advance for any help.

Christer Thrane

************************************************
Professor Christer Thrane
Department of Social Science
Lillehammer University College
2626 Lillehammer, Norway
+ 47 61 28 81 70 (fax)
+ 47 61 28 82 47 (phone, work)
+ 47 61 25 53 04 (phone, home)
E-mail, work: christer.thrane@hil.no
E-mail, home: ch-thran@online.no
************************************************


*
*   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/