Glimmix for ordinal data

[PaigeMiller]

This works for me. You have to use LINK=GLOGIT and not DIST=MULTINOMIAL

proc glimmix data=sashelp.cars(where=(type='Sports'));
ods output parameterestimates=parms1;
nloptions gconv=0.01;
class cylinders;
model origin(ref='Asia')  = cylinders msrp / solution link=glogit;
run;
ods trace off;
proc glimmix data=sashelp.cars(where=(type='Sports'));
ods output parameterestimates=parms2;
class cylinders;
model origin(ref='USA') = cylinders msrp / solution link=glogit;
nloptions gconv=0.01;
run;
proc glimmix data=sashelp.cars(where=(type='Sports'));
ods output parameterestimates=parms2;
class cylinders;
model origin(ref='Europe') = cylinders msrp / solution link=glogit;
nloptions gconv=0.01;
run;

[mgx]

I need the cumulative probit link for my model. I noticed the following error in the log: 

Hence, I think it is not possible in SAS.

[PaigeMiller]

---

@mgx wrote:

I need the cumulative probit link for my model. I noticed the following error in the log: 

---

As far as I know, this is correct. The REF= in the MODEL statement works only for binary or generalized logit models. But even if you don't like the choice of REF=, the predicted values and all other model statistics are correct, and the relationship between the response intercepts is also correctly computed.

[SteveDenham]

I can't find the original post, but I don't see any reason that a cumulative problit link would not be valid with DIST=MULTINOMIAL. Note that for either cumulative link, the first category is assumed to be the reference in this parameterization. If you want to compare somehow to a different level as a reference and still use a cumulative link, you will probably have to use an LSMEANS statement with the diff option, or (much better) use an LSMESTIMATE statement. Comparisons will be on the linked scale. If you want comparisons on the original scale, you ought to look into the NLmeans and NLest macros.  @StatDave has posted many times about this.

However, since I don't see the original post, and @PaigeMiller 's response refers to the use of a generalized logit, I think that maybe your data does not really reflect an ordinal scale, which would then require the use of a reference category.

SteveDenham

[PaigeMiller]

I don't know what happened, but somehow when I replied to an earlier thread, it showed up as a new thread, but with title "Re: Glimmix for ordinal data", the "Re:" indicating it was intended as a reply. Perhaps the forum has a poltergeist.