12-13-2013 06:22 AM
I got the inverse Mills ratio (imr) for all the observations from the proc qlim; however I am unable to reconcile these estimates for imr with the manual estimation where
imr=pdf('normal',pred_xb)/(1-CDF('normal',pred_xb)), if dummy=1 and
imr=-pdf('normal',pred_xb)/(cdf('normal',pred_xb), if dummy=0
qlim procedure estimates and the above estimates are different, I am not sure why, any ideas?
Another issue in relation to proc qlim, why Lambda and the coefficients on Lambda do not show on the outcome of the second equation?