Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Firat Yaman <firat80@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: asclogit collinear problem |
Date | Fri, 19 Feb 2010 11:05:00 +0100 |
I am estimating a regional choice model. I have 5500 cases and 104 regions, but keep only 50 (including the chosen alternative) of them randomly because of memory. When I run asclogit with immigrant concentration as the only covariate, the command works fine. As soon as I add another alternative specific variable (such as population, average house prices etc) I get a warning "model has collinear variables, convergence may not be achieved". Indeed, the loglikelihood function appears to be non-concave. Note that I am not using any variables in casevars(), and that McFadden proves the concavity of the LL function. Can anyone help me out?? * * 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/