Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Hausman test -- what's wrong here


From   "Gustavo Sanchez" <gsanchez@stata.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Hausman test -- what's wrong here
Date   Wed, 11 May 2005 12:36:21 -0500

Dongshu asked,

>    I have two questions:

>   1- If I run this:
>   ivreg Y (X1= A1 A2) X2
>   And I am doing Hausman test to see if the OLS and IV are different.
>   I compare it with
>  OLS1 : reg Y X1 X2

>  my question is that shall I also compare OLS2 :reg Y X1 X2 A1 A2 ?
>  is it meaningful than OLS1

	In order to determine what is the adequate estimator that should be used
you
	don't need to perform any comparison with OLS2.

Next Dongshu asked,

>   it appeared two columns, normally i think the first one (showed as
>"b" in the stata output ) is e1, and then comes e2 (showed as "B" in
>the output) , however,
>it say "b= consistent under Ho and Ha; obtained from regress" and "B=
>inconsistent under Ha, efficient under H0, obtained from ivreg"
>  I am confused here, why it says b is from regress while e1 was from
>my IV regression? the same like B. why it says the opposite?

	This is a bug that was already fixed. While connected to Internet type,
	"update all" and follow the instructions. You will get the correct output
	after updating Stata.

Dongshu also asked:

>  Also, I found that if I performed
>   reg Y X1 X2 A1 A2
>  est store e3
>   hausman e1 e3

>  While the hausman test on e1 and e2 gave me positive number, the
>latter hausman (e1 and e3) comes along with negative test-statistics.
>Why?

As I said before, you shouldn't perform the Hausman test using the
specification
in this last example. However, the result that you are getting may occur
even with
a well-specified problem. The reason for getting a negative test-statistic
is that
the hausman test relies on asymptotic properties; with an infinitely large
sample
the difference between the two covariance matrices is guaranteed to be
positive
definite, but with finite samples that need not be the case.

Gustavo
gas@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