Interpreting regression estimates in PROC GENMOD with Tweedie distribution

[MigMaster12]

Hello all,

I have a question regarding the following PROC GENMOD example in the attached screenshot, using a Tweedie distribution & a log link, found in this SAS documentation: http://support.sas.com/kb/60/335.html

My question is: how would one go about interpreting the regression beta estimates? 

Would it be similar to a regression with a log link? For example: CrRuns yields an estimate of 0.0017, and so the interpretation would be something like "1 unit increase in CrRuns results in exp(0.0017) times the outcome variable salary"?

Thank you again for your help.

[Ksharp · Accepted solution]

Yes. I think so.

log(Yx+1)-log(Yx)=0.0017

=>

Yx+1/Yx = e^0.0017