topic Re: How to perform mixed effect ordinal logistic regression? in Statistical Procedures

How to perform mixed effect ordinal logistic regression?

Diels_O — Sun, 04 Dec 2022 13:23:42 GMT

Dear all, I hava a question about the proc glimmix; the following was my code

 proc glimmix data=pt empirical order = data PLOTs= all;
 class Affiliation pt60dich3(ref = first)/ref = first;
 model score = pt60dich3 / dist=bin link=glogit SOLUTION cl oddsratio;
 random intercept/subject = Affiliation; run;

Here, score was a ordinal score.

However, this model went wrong: the generalized logit can only be used with the multinomial distribution.

How should i perfrom mixed effect ordinal logistic model?

Re: How to perform mixed effect ordinal logistic regression?

PaigeMiller — Sun, 04 Dec 2022 17:30:30 GMT

Ordinal implies (to me) more than two levels of the response variable. Is that what you have?

If so, then BIN is the wrong value for DIST. You do need to specify the MULTINOMIAL distribution.

Re: How to perform mixed effect ordinal logistic regression?

StatDave — Mon, 05 Dec 2022 16:07:01 GMT

The generalized logit model (LINK=GLOGIT) is typically used when the multinomial response is nominal. If your multinomial response is ordinal and you want the model to account for its ordinality, then you can use the cumulative logit model (LINK=CUMLOGIT). The cumulative logit model is the default when you specify DIST=MULT.

Re: How to perform mixed effect ordinal logistic regression?

Diels_O — Mon, 05 Dec 2022 19:25:30 GMT

Thank you for your quick response!