topic Proc genmod - Response variable exponentially distributed in Statistical Procedures

Proc genmod - Response variable exponentially distributed

FranciscoB — Thu, 10 Sep 2009 17:07:46 GMT

Hi everyone,

I am currently trying to fit a GLM using proc genmod and my response variable is exponentially distributed. I can't seem to find how to set this up since the exponential distribution is not among the options for dist.

If it is not possible, I would like to ask then if it's possible to fix one of the parameters of a Gamma distribution so that I can fix the response variable distyribution as Gamma(1,mean).

Does anyone have any suggestion about what to do when most of the data (supposedly continuous) has value 0?

Thank you in advance

Re: Proc genmod - Response variable exponentially distributed

Dale — Fri, 11 Sep 2009 20:35:38 GMT

Perhaps you want to use the LIFEREG procedure instead of PROC GENMOD. The LIFEREG procedure allows specification on the MODEL statement that the distribution is exponential.

Dale

Re: Proc genmod - Response variable exponentially distributed

StatDave — Mon, 14 Sep 2009 19:30:44 GMT

The exponential distribution is obtained when the scale parameter of the gamma distribution (nu in the GENMOD documentation) is 1. So, you can fit a model to exponential data by simply adding the SCALE=1 and NOSCALE options in the MODEL statement. The NOSCALE option keeps the scale parameter fixed at the SCALE=1 setting and gives you a Lagrange multiplier test of the nu=1 restriction.