Re: st: mixlogit log normal distribution restriction problems

[Arne Risa Hole <arnehole@gmail.com>]

Hi Sorada,

You can get around this by supplying your own starting values using
the -from- option of -mixlogit- (see -help maximize-). By default
-mixlogit- uses the coefficients from a -clogit- model estimated on
the same data as starting values, but if these are negative it issues
the error msg that you reported.

I hope this helps.

Arne (author of -mixlogit-)

2010/2/26 Sorada Tapsuwan <sorada_tapsuwan@hotmail.com>:
>
> Hi
> I'm trying to estimate a mixlogit model and restrict a number of random variables to follow a lognormal distribution using the command ln(#).
> However, when I ran the model it gave me an error message that says "Variables specified to have log-normally distributed coefficients should have positive coefficients in the conditional logit model. Try multiplying the variable by -1."
> I have about 4 variables that have -ve coefficient values in the clogit model but none of them were significantly different from zero, so it's not like the coefficient is definitely -ve. So I figured that if I try to estimate them using mixlogit and restrict them to follow a lognormal distribution they should be okay but STATA doesn't like it. How can I then restrict these 4 variables to follow a lognormal distribution then?
> Would anyone be able to help me with this problem. How can I overcome this constraint?
> Thanks
> S
>
> _________________________________________________________________
> Hotmail: Trusted email with Microsoft’s powerful SPAM protection.
> https://signup.live.com/signup.aspx?id=60969
> *
> *   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/

*
*   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/