Page 1 of 1
Wald test between estimated results of two regressions?
Posted: Thu May 14, 2015 3:38 pm
by Aixia_Mei
Hi Tom,
I run the same program with pre-event and post-event data separately, thus the estimated results of each coefficients are in two different files. My question is how can I apply Wald tests on the equality of pre- and post- estimated coefficients ?
Many thanks,
Aixia
Re: Wald test between estimated results of two regressions?
Posted: Thu May 14, 2015 4:57 pm
by TomDoan
Assuming that you're willing to assume the two estimates are statistically independent, then the covariance matrix of the difference is the *sum* of the separately estimated covariance matrices. The calculation is very similar to that with a Hausman test except with the Hausman test the covariance matrix is the *difference* of the more efficient from the less efficient. (In the HT, the estimates are very highly correlated rather than being independent). The grnp173.rpf example shows how to do the calculation.
Regarding the fact that the two are based upon different data files, is there any reason you can't read in both data sets into one program? That's the simplest way to handle it. Otherwise, you would have to do a WRITE instruction from one program to send %BETA and the covariance matrix to a file, then a READ instruction in the other to read them in for the further calculation.