Re: st: FW: comparing xtregar coefficients across models

[Christopher Baum <kit.baum@bc.edu>]

<>
On Feb 26, 2011, at 2:33 AM, Lopa wrote:

> 
> The problem was that I was trying to test the equality of the two t-stats. Instead, I needed to use the coefficients generated by the two models for the test command.
> 
> Run model 1.
> Save the beta in a scalar as:
> . scalar b_lagseaavgwqfoia04avg3=_b[lagseaavgwqfoia04avg3]
> Run model 2.
> Run test as:
> . test _b[pastyearseaavgwq]=b_lagseaavgwqfoia04avg3

This makes no sense, and does not address the underlying problem of statistical inference. You are testing an estimated coefficient from the second model against a fixed number. That test assumes that the number from the first model is your hypothesized value of the coefficient, and completely ignores the sampling error associated with that coefficient estimate. The test you have run would make sense if the fixed number was some theory's prediction about that coefficient.

I don't know if -suest- works in your context, but here is a valid test of a similar hypothesis with -regress-:

sysuse auto,clear
qui reg price mpg turn if foreign
est sto f
qui reg price mpg turn if !foreign
est sto d
suest f d
test [f_mean]turn = [d_mean]turn


Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html




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