topic Re: The reasoning for regression node's class reference group in SAS Data Science

The reasoning for regression node's class reference group

paradigm_fellow — Thu, 09 Apr 2015 19:22:45 GMT

Hi everyone,

I know that the regression node in SAS Enterprise Miner does not allow us to change the class references groups. However, EM picks a references group for you. What is the reasoning behind picking a specific group for a class variable?

Just to let everyone know, I am running a series of logistic regression and I am using EM 13.2.

Re: The reasoning for regression node's class reference group

M_Maldonado — Fri, 10 Apr 2015 17:26:21 GMT

Have you tried the HPGLM node?

You can set Reference Level as User Defined and then use Reference Level menu to set choose the reference level.

I hope it helps,

Miguel

Re: The reasoning for regression node's class reference group

paradigm_fellow — Fri, 10 Apr 2015 20:00:18 GMT

Hi Miguel,

Thank you for the reply and the help. I was able to do it on HPGLM. However, I now have a new problem. It is not giving me odds ratios. Is there a way to get the odds ratios in HPGLM?

Respectfully,

Frank

Re: The reasoning for regression node's class reference group

Funda_SAS — Fri, 10 Apr 2015 20:51:56 GMT

Hello Frank,

If your goal is to get the odds ratios, I recommend you to use PROC LOGISTIC. This PROC also enables you to change reference levels for your class variables.

The reason why HPGLM node does not provide odds ratios is that it runs on the HPGENSELECT procedure. High-performance statistical procedures (procs start with HP) and EM nodes are designed for predictive modeling and other large-data tasks, they do not implement all of the inferential features that are available in corresponding traditional SAS procedures.

Hope you find this useful!

Funda