Re: st: ML convergence with Stata 12 and Stata11

[jpitblado@stata.com (Jeff Pitblado, StataCorp LP)]

Maria Yanotti <Maria.Yanotti@utas.edu.au> is using -logit-/-probit- in Stata
11 on a 32-bit laptop and Stata 12 on a 64-bit desktop, using the same data
and model specification; Stata 11 converges to a solution but Stata 12 does
not:

> I currently have Stata/SE 11.2 for Windows (32-bit) in my laptop and I am
> running Stata/MP 12.1 for Windows (64-bit) in my desktop. I am estimating a
> probit/logit model with over half a million observations. When I run the
> probit model with maximum likelihood (ML) in Stata/SE 11.2 in my laptop,
> after 6 iterations and 11 seconds I get an estimation (the log
> pseudolikelihood converges to -367498.27); when I run exactly the same model
> with the same data using ML in Stata/MP 12.1 in my desktop the ML gets stuck
> in iteration  -367496.9(backed up), and it takes over 10000 iterations or
> more to get an output (which may not converge) and these take more than 3
> days.
> 
> The commands I use are:
> 
> eststo Prob1: probit depvar indvars, vce(r)
> 
> Can anyone please help me understand what is going on? How can I solve this
> problem? I am not able to use my powerful desktop for my estimations and end
> up estimating everything on my laptop with very low memory.

There are no real differences in the code for -probit- (or -logit-) between
Stata 11 and Stata 12, so I suspect the difference is explained by the two
different machines.  We recommend that Maria email tech-support

	http://stata.com/support/tech-support/contact/

and include a simple do-file with a dataset that reproduces the problem.

--Jeff
jpitblado@stata.com
*
*   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/