st: Nested logit, individual-level variables

["Lubell, Mark" <mnlubell@ucdavis.edu>]

Hi STATA list:

This question has been asked in the past, but I really have not seen a sufficient answer on the archives.  I trying to run nested logit but only with individual-level variables, for example, education and income.  The survey respondents are choosing among 4 alternatives, and I have created a that has one nest with only one alternative, and another nest with 3 alternatives.  I have created dummy variable interactions with the individual-level variables in order to get variance in the groups/cases.  However, the MLE appears not to be converging for this model.  I'm wondering if there is a theoretical reason that nested logit cannot handle individual-level data only, for example in multinomial probit there are parameters that are unidentified without alternative-specific variables.

Any advice appreciated on whether or not nested logit is capable of being run with individual-level only data, and if so, how to make it work.

Thanks, Mark

Mark Lubell, Ph.D.
Department of Environmental Science and Policy
One Shields Avenue
University of California, Davis
Davis, CA 95618

Email: mnlubell@ucdavis.edu
Center for Environmental Policy and Behavior:  http://environmentalpolicy.ucdavis.edu/
Personal Homepage:  http://www.des.ucdavis.edu/faculty/lubell/
Phone:  530-752-5880
Fax: 530-752-3350
Office: 2146 Wickson Hall


*
*   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/