topic Re: Proportional odds assumption in proc surveylogistic in Statistical Procedures

Proportional odds assumption in proc surveylogistic

ChuksManuel — Fri, 18 Sep 2020 04:02:02 GMT

Hello programmers,

I am trying to test for proportional odds assumption using sas proc surveylogistic. It seems I am able to do this using proc logistic. But I want to account for the complex sampling in my data, so surveylogistic is a given. Is there a way proc survey logistic implements this? I am doing an ordinal logistic regression.

proc logistic data=asth ;
class MDD (Ref= '0') mold (Ref= '0') age_new (Ref= '1') smoker(Ref= '0') gender (ref='0')  BMI (ref='0') / param=ref;
model activity (descending) = MDD mold age_new smoker gender BMI ;
run;

proc surveylogistic data=asth;
strata strt ;
cluster psu ;
weight wgt;
class MDD (Ref= '0') mold (Ref= '0') age_new (Ref= '1') smoker(Ref= '0') gender (ref='0')  BMI (ref='0') / param=ref;
model activity (descending) = MDD mold age_new smoker gender BMI ;
run;

Re: Proportional odds assumption in proc surveylogistic

ChuksManuel — Fri, 18 Sep 2020 04:06:42 GMT

And i tried the template code in the SAS Proc surveylogistic handbook https://support.sas.com/documentation/onlinedoc/stat/131/surveylogistic.pdf. And it did not give me the proportional odds assumption test in the output. Did this change?

Re: Proportional odds assumption in proc surveylogistic

SAS_Rob — Fri, 18 Sep 2020 11:51:34 GMT

In previous releases SAS reported a test of the Proportional Odds Model in SURVEYLOGISTIC. But there are some serious theoretical questions that have been raised about using the Test for Proportional Odds for survey data that have brought into question how and even if this should be reported for survey data. Until those issues are resolved, SAS removed it from the output. At the very least it should be using the survey adjusted variance and not the inverse of the information matrix the way LOGISTIC does.

Re: Proportional odds assumption in proc surveylogistic

ChuksManuel — Fri, 18 Sep 2020 13:57:16 GMT

Thank you for the response!

Re: Proportional odds assumption in proc surveylogistic

bherbert — Wed, 07 Apr 2021 19:10:26 GMT

Could you please provide a reference for this?