Re: st: constrained parameter in maximum likelihood estimation

[eivind.bernhardsen@econ.uio.no]

Dear Jun Xu,

I do have an alternative suggestion:

If your reason for constraing the variable is to keep the likelihood 
function from settling in some local maximum outside the [0,1] range (that 
is you believe the global maximum can be found within this range), you can 
try to replace your parameter p with 1-(sqrt(1-(sqrt(p))^2))^2 at some 
place in your likelihood program. This fuction will return p in the [0,1] 
range and "." outside this range and thus keep the solver from iterating 
outside [0,1].

(Presumably you should use this one with care...)


Regards,
Eivind Bernhardsen


At 11:44 24.09.2003 -0500, you wrote:

> Stata Listers:
>
> I have a programming question using ml command: how to contrain some 
> parameter to be estimated within certain range, for example, some 
> probabilities [0,1]. One way that I can think of is to do some logit 
> transformation like exp(p)/exp(1+exp(p)) and let ml freely estimates p. 
> However, it is hard to get its standard error except using the delta 
> method, which is not very accurate. Is there way in Stata, I can constrain 
> the value of p, while at the same time direct get the asymptotic standard 
> error without using any transformation or delta method? Thanks a lot.
>
> Jun Xu
> Doctoral Student
> Department of Sociology
> Indiana University at Bloomington
>
> _________________________________________________________________
> Get McAfee virus scanning and cleaning of incoming attachments.  Get 
> Hotmail Extra Storage!   http://join.msn.com/?PAGE=features/es
>
> *
> *   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/

*
*   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/