Renzo,
> program define mylikelihood
> args lnf theta1 theta2
> replace
> `lnf'=37*ln(norm((1000-`theta1')/`theta2')+30*ln(1-norm((1000-`theta1')/
> `theta2'))
> end
>
>
> ml model lf mylikelihood (Rincome= LB marital educ race xnorcsiz)
> ml check
> ml search
> ml maximize
> ml clear
>
>
> Any idea? Suggestions?
It would help a lot if you provided the output and the error messages you
got.
Off the top of my head, I see a couple of difficulties with your program.
The first one is that you have two `theta's, but only specify one equation
in your -ml model- statement -- you would need to add another one which
would probably look like /sigma=.
Another thing seems more substantial -- are you sure your model is
identified? My understanding of probit regressions has been that the
variance of residuals is standardized to one, and not without a reason:
otherwise, you can scale up both the regression coefficients and the
variance by a common factor without changing the model and the likelihood.
You might have this problem if you have a constant term among your
regressors. Stata will not be able to find all this out by itself, but you
won't make it converge. That's exactly what lack of identification is:
there are many values of the parameters which yield exactly the same
likelihood, i.e., there is a ridge in the likelihood profile, and Stata
would have tremendously difficult time around it (even with -difficult-
option).
--- Stas Kolenikov
-- Ph.D. student in Statistics at UNC-Chapel Hill
- http://www.komkon.org/~tacik/ -- [email protected]
* This e-mail and all attachments to it are not intended to provide any
* reasonable point of view and was transmitted to you in error. It
* should be immediately deleted by all recepients unless they really
* enjoy communicating with the author :). Other restrictions apply.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
