topic Re: Proc GLIMMIX in SAS Procedures

Proc GLIMMIX

zen_cool — Sat, 06 Apr 2013 16:18:29 GMT

The following SAS code is from http://www.ats.ucla.edu/stat/sas/faq/anovmix1.htm

PROC MIXED DATA=long;
CLASS sub group trial;
MODEL dv = group trial group*trial;
REPEATED trial / SUBJECT=sub TYPE=CS;
run;

I was just wondering how you would use PROC GLIMMIX to write this model. Would it be:

PROC GLIMMIX DATA =long;

class sub group trial;

MODEL dv = group;

RANDOM trial/ SUBJECT = sub TYPE = CS;

RANDOM trial sub trial*sub trial*group;

Also out of curiosity, would the model be as follows:

DV = GROUP + TRIAL + SUB + TRIAL*SUB + TRIAL*GROUP where the terms after GROUP are all random?

Re: Proc GLIMMIX

lvm — Mon, 08 Apr 2013 16:25:30 GMT

No, your glimmix code is fitting a different model. Just take your MIXED code and change your REPEATED statement to:

RANDOM TRIAL / SUBJECT=SUB RESIDUAL TYPE=CS;

If you do not have missing values for trial (i.e., same trials for each subject), you can use:

RANDOM _RESIDUAL_ / SUBJECT=SUB TYPE=CS;

Re: Proc GLIMMIX

zen_cool — Tue, 09 Apr 2013 01:06:36 GMT

What does the RESIDUAL option do?

Re: Proc GLIMMIX

SteveDenham — Tue, 09 Apr 2013 12:10:04 GMT

It sets up the R side matrix (repeated).

Steve Denham

Re: Proc GLIMMIX

lvm — Tue, 09 Apr 2013 13:23:42 GMT

As Steve says, the statements in GLIMMIX

random _residual_ / sub=___ type=____;

random ___ / sub=____ type=____ RESIDUAL;

do the same thing as the REPEATED statement in MIXED. Since you had a model in MIXED that you wanted to fit in GLIMMIX, I showed you the appropriate code.