topic Re: Estimate statement in proc logistic for multinomial outcome in Statistical Procedures

Estimate statement in proc logistic for multinomial outcome

useSAS — Mon, 10 Jul 2017 21:44:14 GMT

I have some trouble in using estimate statement in PROC logistic for multinomial outcome. Is there any example for this situation ?

Thanks

Re: Estimate statement in proc logistic for multinomial outcome

Reeza — Mon, 10 Jul 2017 22:03:07 GMT

@useSAS wrote:

I have some trouble in using estimate statement in PROC logistic for multinomial outcome. Is there any example for this situation ?

Thanks

Please post your question directly in the forum, not as an attachment, as many users will not open them.

Additionally, remove your Site Number from your post.

Re: Estimate statement in proc logistic for multinomial outcome

StatDave — Tue, 11 Jul 2017 18:38:55 GMT

This cannot be done with the ESTIMATE statement. But you can use the general NLEstimate macro to test any linear or nonlinear combination of model parameters. This note shows one example of using the macro with a multinomial model. Links to other examples of using the macro are given in the Results tab of the macro description. For your example, you just want to contrast the 3rd and 4th parameters in the model. That is, you want to test the hypothesis that the difference in those parameters is zero. Here is a similar example using the Crops data in the example titled "Scoring Data Sets" in the LOGISTIC documentation. Include a STORE statement to save the fitted model information in an item store named LOG. After submitting the macro definition code to define the macro (see the Usage section of the macro description), call the macro and specify the function to estimate using the parameter names that the macro assigns: B_p1, B_p2, ... .

proc logistic data=crops;
where crop in ('Corn','Soybeans','Cotton');
model crop = x1 / link=glogit;
test x1_Corn=x1_Cotton;
store log;
run;
%NLEstimate(instore=log, label=Diff, f=B_p3-B_p4, df=15)

Re: Estimate statement in proc logistic for multinomial outcome

useSAS — Thu, 13 Jul 2017 01:33:54 GMT

Thank you !