04-19-2018 02:38 PM
It's not clear to me whether you have covariates or not. There is an example of fitting a gamma-distributed response in the PROC GENMOD documentation. For the intercept-only example with a log link, the doc says "the intercept is the estimated log mean of the fitted gamma distribution so that the mean ...is mu = exp(Intercept)."
For general models, the mean is conditional on the covariates (through the link function). For a particular choice of covariates , the conditional mean is mu(x) = inverse_link(x`*beta ).
04-19-2018 03:12 PM
The predicted values available using the PRED= option in the OUTPUT statement provide the estimate of the mean (mu) in the population defined by each observation.