Re: st: Bootstrap with version control

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Thu, 14/10/10, Dmitriy Poznyak wrote:
> I am trying to fit an -mlogit- model based on survey data
> in version 11. Since some of my predictors are categorical
> and entered as dummies, Spost post-estimation doesn't seem
> to work with version 11, but runs smoothly under version
> control (v. 10.1). However, when I bootstrap the data, which
> I need for some reasons, I get an error in v. 10.1. The same
> model converges smoothly in v. 11.
> My set-up is as follows:
> 
> 1) bootstrap, reps(100) force : mlogit y x z... , cl(nis),
> [pweight=weight1] > this one runs smoothly but no
> prchange, prvalue, etc. can be obtained as I am using
> categorical predictors

You don't have to use factor variable notation in Stata 11,
you can still create your own dummies as we used to do in
Stata 10. If you do that, then you don't need to use version
control and there is no problem with the user writen programs 
like -prchange-, -prvalue-. I guess that that is the easiest
solution.
 
> 2) version 10.1: bootstrap, reps(100) force : mlogit y x
> z... , cl(nis), [pweight=weight1] this one doesn't converge,
> although when bootstrap is opted out, prtab and prchange
> (not prvalue though) work just fine.
> 
> I couldn't find whether bootstrap works w/ version control
> but I can't see any reason why it shouldn't.

The exact program/algorithm used in finding the maximum of
the likelihood function changed. Normally this does not
matter, but it can make a difference if you are estimating 
a model that is (almost) too complex for your data.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------




      

*
*   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/