07-04-2017 10:28 AM
I am running a tobit regression and want to report overall (mean) marginal effects and the standard errors of the overall marginal effects.
My dependent variable is earnings which cannot be negative and my explanatory variables are treatment (dummy) and some covariates.
I use the following code:
proq qlim data=Sample;
Model earnings=treatment age gender;
output out=est marginal;
endogenous earnings ~ censored (lb=0);
proc means data=est mean stderr;
var meff_treatment meff_age meff_gender;
The problem is that the std.errors and correspondng p-values in the qlim procedure are boarderline insignificant while the std.errors from the proc means procedure are higly significant (p<0.001) so I guess it cannot be done this way. Is there some other way to do this?