Re: st: how have stata perform and report a likelihood-ratio test after running biprobit

[Nick Cox <njcoxstata@gmail.com>]

Referring to "another computer" or what you "read somewhere" is not
informative.

You need a report of the following kind

1. Data we can see or use or download to check what you did.

2. Precisely which Stata was used.

3. Precise results that differ.

Note that upgrades of Stata sometimes withdraw functionality
previously implemented if theoretical arguments and/or extensive
simulation testing lead StataCorp to believe that some detail in the
previous code was a bad idea.

Nick

On Tue, Dec 18, 2012 at 3:04 PM,  <Xing.Liu@naturvardsverket.se> wrote:

> Thanks for your quick relpy. But however, when I run the exactly same model (BIPROBIT), using the exactly same data on another computer, it does give me likelihood ratio test. And I read somewhere that likelihood ratio test is better since Wald test requires large sample size but I only have around 300.
>
> And Wald test gives that rho is not significantly different from zero, and likelihood ratio test gives that rho is significant different from zero. I thought I should trust the likelihood ratio test.

Maarten Buis

> You should not do so, as it typically means that the likelihood ratio test is not appropriate in that situation, see:
> <http://www.stata.com/support/faqs/statistics/likelihood-ratio-test/>

On Tue, Dec 18, 2012 at 12:53 PM,  <Xing.Liu@naturvardsverket.se> wrote:
>> I noticed that sometimes stata reports Wald test for rho=0 below the table after running biprobit, sometimes it reports likelihood-ratio test below the table after running biprobit.
>> I am wondering how to have stata perform likelihood ratio testafter biprobit when it chooses to perform Wald test?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/