 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

Re: st: A test for variation in regression coefficients

 From "Michael N. Mitchell" To statalist@hsphsun2.harvard.edu Subject Re: st: A test for variation in regression coefficients Date Fri, 22 Oct 2010 20:24:18 -0700

Dear Jeff

I wonder if this example does the trick for you...

sysuse auto, clear
rename price y
rename trunk x2
rename weight x3
rename length x4
rename turn x5

regress y x1 x2 x3 x4 x5
test x1=x2
test x2=x3, accum
test x3=x4, accum
test x4=x5, accum

So the final test yields...

( 1)  x1 - x2 = 0
( 2)  x2 - x3 = 0
( 3)  x3 - x4 = 0
( 4)  x4 - x5 = 0

F(  4,    68) =    4.13
Prob > F =    0.0047

Which is the test that B1 = B2 = B3 = B4 = B5, which has 4 df and in this case is significant.

Is that what you are looking for?

Michael N. Mitchell
Data Management Using Stata      - http://www.stata.com/bookstore/dmus.html
A Visual Guide to Stata Graphics - http://www.stata.com/bookstore/vgsg.html
Stata tidbit of the week         - http://www.MichaelNormanMitchell.com

On 2010-10-22 7.52 PM, Jeff Roberts wrote:
Hi,

I am interested in a test for variation in regression coefficients.  I have
a block of coefficients (20 in all) for which I’d like to show that that the
coefficient estimates vary one from another.  This would be test of the null
hypothesis that Beta(i) is constant for all i.
I have constructed linear contrasts to test that each coefficient is
different from the “average” of the remaining coefficients; however, this
not really the single test statistic that I’m after.
Any thoughts would be greatly appreciated.

Thanks and Best Regards,
J.Roberts

*
*   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/
*
*   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/