topic A question about the PROC GENMOD in Statistical Procedures

A question about the PROC GENMOD

John4 — Wed, 28 Jun 2017 10:21:45 GMT

I'am using the PROC GENMOD like this :

PROC GENMOD DATA = LIFDAT;
CLASS MFG;
MODEL LIFETIME = MFG / DIST=GAMMA LINK=LOG TYPE3;
RUN;

but I dont understand why an estimated parameter equals zero

What does it means ? can we change that ?

Re: A question about the PROC GENMOD

Rick_SAS — Wed, 28 Jun 2017 11:46:52 GMT

See "Parameterization of Model Effects" in the SAS documentation.

When you use a classification variable, one of the levels is redundant. In the GLM parameterization (the default for PROC GENMOD) the parameter for that level is set to zero.

Re: A question about the PROC GENMOD

PaigeMiller — Wed, 28 Jun 2017 12:08:58 GMT

In the CLASS statement you can add an option to change the parameterization of your model to the EFFECT, this may be what you are looking for.

class mfg/param=effect;

but please note the resulting model is the same, and the predictions are the same, just the way the model is parameterized has changed.

In this situation, you may be more interested in the LSMEANS, which sound like what you really want to look at here instead of the model effects, and the LSMEANS don't change depending on the parameterization of the model.