1
$\begingroup$

I have just done a Chow test on a regression in order to see whether there is a structural break. I am a bit stumped however as my Chow test returns a negative number. Now what do I do?

More specifically, this expression (from the Wikipedia entry for the Chow test):

$\frac{(S_c-(S_1+S_2))/(k)}{(S_1+S_2)/(N_1+N_2-2k)}$

turns out negative for me.

EDIT: It turns out it was a simple programming error.

$\endgroup$
  • $\begingroup$ Does your regression has an intercept? Omitting intercept sometimes can cause problems. Also could you post reproducible example? It is clear that you made an error, as @onestop pointed out, but with example available it may be possible to point out where the error is exactly. $\endgroup$ – mpiktas Feb 28 '11 at 12:29
  • $\begingroup$ Cheers! Thanks for your edit. Simple programming errors are the biggest enemy of scientific progress :) $\endgroup$ – mpiktas Apr 7 '12 at 20:29
4
$\begingroup$

You've made a mistake somewhere in your calculations. It's not possible for the sum of the squared residuals from a single regression using the combined data to be less than the sum of the sums of squared residuals from the regressions using the two separate sets of data.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.