02-24-2014 05:57 PM
I'm doing this kind of regression:
proc genmod data= data;
class id year / param=ref;
x v w z
/ d=nb dscale;
estimate 'x' x 1 / exp;
estimate 'z' z 1 / exp;
Since the nb is not a linear function I can not directly compere the coefficients (i.e betaX and betaZ) but I need to 1) look at the IRR and 2) to implement a post estimation analysis to test whether the coefficients estimated for x and z, are actually statistically different
The story in stata it is very easy: after your reg (for ex nbreg y x v w z, robust) the code is: test x= z
I'm trying to implement the same analysis in SAS with the regression above. I tried to use ods output and to implement ttest but I did get any results.
Actually my knowledge of SAS is not good as well my knowledge in econometrics and I did not find any clear description on internet about that test.
Can you help me?