topic Re: Proc Logistic v. HP Logistic Certain Output Predicted Probabilities Multinomial Case (SAS EG 7.1 in Statistical Procedures

Proc Logistic v. HP Logistic Certain Output Predicted Probabilities Multinomial Case (SAS EG 7.1)

davis — Wed, 22 Sep 2021 19:00:01 GMT

Good afternoon,

I have a question, and was wondering if someone could either lend a hand or point me in the right direction.

I conducted a multinomial run using proc logistic. One output is the predicted probability for each outcome in the following form:

Pred_A=exb_A /(1+exb_A+exb_B+exb_C+exb_D);

Pred_B=exb_B /(1+exb_A+exb_B+exb_C+exb_D);

Pred_C=exb_C /(1+exb_A+exb_B+exb_C+exb_D);

Pred_D=exb_D /(1+exb_A+exb_B+exb_C+exb_D);

Pred_E=1 /(1+exb_A+exb_B+exb_C+exb_D);

where Pred_E = 1/(1+Pred_A+Pred_B+Pred_C+Pred_D) and produced automatically by proc logistic.

Does HP Logistic have a similar feature for a multinomial run?

Thanks.

Re: Proc Logistic v. HP Logistic Certain Output Predicted Probabilities Multinomial Case (SAS EG 7.1

sbxkoenk — Wed, 22 Sep 2021 19:20:50 GMT

Hello,

For PROC HPLOGISTIC contrasted with PROC LOGISTIC, look here:

https://go.documentation.sas.com/doc/en/statug/15.2/statug_hplogistic_overview02.htm

And for multinomial response, use LINK=GLOGIT/GENLOGIT.

If you specify LINK=GLOGIT, the procedure assumes a generalized logit model for nominal (unordered) data, regardless of the number of response categories.

https://go.documentation.sas.com/doc/en/statug/15.2/statug_hplogistic_syntax07.htm#statug.hplogistic.model_link

Then you get similar output to PROC LOGISTIC, no?

If not, post your PROC LOGISTIC code and we will reply with the equivalent PROC HPLOGISTIC code.

Regards,

Koen