Hi all,

I was wondering if there is a way to make meaningful inferences on the coefficients that are estimated by reg3 based on two sub-samples? Assume the following:

Code:
reg3 (winv2 wndf wdiv wsize wcf wmtb cage wtang l.stdev l.winv2 i.sectorno) (wndf winv2 wdiv wmtb wcf wtang wsize wol l.stdev i.sectorno) if ablev>0
reg3 (winv2 wndf wdiv wsize wcf wmtb cage wtang l.stdev l.winv2 i.sectorno) (wndf winv2 wdiv wmtb wcf wtang wsize wol l.stdev i.sectorno) if ablev<0
I used Hausman test and rejected the null hypothesis that difference in coefficients are not systematic for each of the equations. But can we compare the actual impacts of coefficients on the basis of one standard deviation? I thought it didn't make sense because one standard deviation in one sub-sample isn't equal to one standard deviation in the other sample.

Is there a way to compare these coefficients by reporting actual number?

Thanks