Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: suest after clogit


From   jpitblado@stata.com (Jeff Pitblado, StataCorp LP)
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: suest after clogit
Date   Tue, 22 May 2007 17:35:06 -0500

Xiaoheng Zhang (Kevin) <zhangx@tcd.ie> is using -suest- with -clogit-:

> I tried hausman test for IIA(Independence from Irrelevant Alternatives) after
> clogit but failed because that chi2<0 means that model fitted on these data
> fails to meet the asymptotic assumptions of the Hausman test.Then Stata refers
> me to the suest.I used following command but at the end Stata propmts that an
> equation cannot be found.
> 
> clogit choice lap lcdrug2 eatr ldist larea lwage3 edu3 share lgi sisters2 if
> location!=6,group(id)
> est store de
> clogit choice lap lcdrug2 eatr ldist larea lwage3 edu3 share lgi
> sisters2,group(id)
> est store fullset
>   suest de fullset,cluster(id)
>   test [de=fullset],common
> equation de not found
> r(303);
>
> My question are that 1) is "test [de=fullset]" the command to do an equality
> test for coefficients of different models?
> 2) Did I make any mistake in the commands I worte?

In Stata 9, -clogit- was partially rewritten to use -ml- for maximizing it's
likelihood (-clogit- was originally an internal command).  As a result, -e(b)-
now contains an equation name, corresponding to the dependent variable.

I believe the -test- command should be:

	. test [de_choice=fullset_choice]

-suest- uses the estimation result's name as part of the new equation name,
the other part comes from the equation name present in the -e(b)-.

> PS:I found these words in help file that "clogit does not have a score() option.
> suest has special code to deal with the case of "one success per group".  If
> multiple clogits are combined, the group() structures should not overlap;
> nested structures are allowed. Also, if cluster() or svy is specified with
> suest, the clustering should not overlap with the grouping in clogit. See [R]
> suest for details."
> 
> Unfortunately,if my understanding is correct, to do a Hausman test for IIA,I
> have to have the same group() for two clogit models to be compared.How could I
> avoid overlapping?

This note was removed as part of the new features in Stata 9.

There is an example very similar to Kevin's in -[R] suest- (bottom of page 310
in the 'R-Z' manual for Stata 9).

--Jeff
jpitblado@stata.com
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index