3
$\begingroup$

I wonder what it means if the statistic (ratio) of this test is one? The statistic is built from sample splitting. Then you have to calculate the ratio RSS2 / RSS1, which are the Residual Sum Squares of the new subsamples.
I think it means that both subsamples have the same variance but I am not sure at all.
Thank you in advance.

$\endgroup$
  • $\begingroup$ Additional information needed: 1) Total number of observations in the sample, 2) number of observation in sub-smaple 2, 3) number of observations in sub-sample 1. Also 4) Number of regressors (including the constant term). $\endgroup$ – Alecos Papadopoulos Jun 12 '15 at 12:12
  • $\begingroup$ And, it would be better to include this information in the question rather than in a comment. $\endgroup$ – Alecos Papadopoulos Jun 12 '15 at 12:26
  • $\begingroup$ It is just a therorical question. Any regression was made in this case. $\endgroup$ – Newbie Jun 12 '15 at 15:19
1
$\begingroup$

The situation implicitly described in the question assumes that the two sub-samples have equal size (which is not necessary for the test to go through).

In such a situation, if the ratio $RSS2/RSS1$ equals unity, then the null-hypothesis of homoskedasticity won't indeed be rejected for any significance level up to $0.5$, (to be compared to the conventional levels of $0.1,\; 0.05,\; 0.01$) because the statistic used, under the null hypothesis follows an F-distribution (exactly or asymptotically), and more over, one that has equal degrees of freedom for the numerator and the denominator (due to the equal sub-sample sizes).

This also makes intuitive sense: if the obtained estimates for the sub-sample variance are equal, then, in order to reject the null, we should accept that it would be more probable than not, that the rejection will be mistaken.

$\endgroup$

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.