topic Re: Help with PROC BGLIMM in Statistical Procedures

Help with PROC BGLIMM

xliu1 — Sat, 21 Mar 2026 20:54:41 GMT

I try to use proc bglimm to predict future data, and I got error message regarding this preddist statement. The combined data includes historical and future data set with missing values assigned to future cohort. Here is a screenshot of the error I see in SAS. I would appreciate if anyone can help me fix this error. Thanks.

proc bglimm data=combined seed=&seed nmc=&nmc thin=&thin nbi=&nbi
	  outpost=Post_Pred;
class global_college;
model PELL_GRAD_6YR / PELL_COHORT_6YR = time FIRST_GEN_PERC SCH90_PERC GPA2_PERC SENIOR_PERC / dist=binomial link=logit;
random intercept time / subject=Global_College type=un;
preddist outpred=Predictive;
title "BGLIMM: Generate Forecasts for Future Cohorts";
run;

Re: Help with PROC BGLIMM

sbxkoenk — Sun, 22 Mar 2026 17:31:44 GMT

There's no ERROR message.
You have just published the SAS-program and not the LOG that you receive after submitting this.

BR, Koen

Re: Help with PROC BGLIMM

xliu1 — Sun, 22 Mar 2026 20:48:46 GMT

Re: Help with PROC BGLIMM

FreelanceReinh — Sun, 22 Mar 2026 22:49:06 GMT

Hello @xliu1,

The PREDDIST statement was introduced in release 2020.1.1 (December 2020) of SAS Viya 4. Apparently, you are using either an older SAS Viya 4 release (like LTS 2020.1), an older Viya version (such as 3.5), or SAS 9.4, where that statement is not supported even in the most recent maintenance release (M9 with SAS/STAT 15.4).

I am sorry that I can't help you further with PROC BGLIMM because my SAS 9.4M5 release (with SAS/STAT 14.3) does not support that procedure at all. It was introduced in SAS 9.4M6, SAS/STAT 15.1.

Re: Help with PROC BGLIMM

xliu1 — Sun, 22 Mar 2026 23:01:12 GMT

Thank you for your response. I am using SAS base 9.4. What options are available for this version of SAS to produce the future predictions using PROC BGLIMM?

Re: Help with PROC BGLIMM

sbxkoenk — Mon, 23 Mar 2026 09:40:55 GMT

The PREDDIST statement is only available in the SAS Viya version of BGLIMM. As mentioned earlier by @FreelanceReinh .

You could also fit the model in PROC MCMC instead which does have a PREDDIST statement.

Here's an example that uses the PREDDIST statement in PROC MCMC (Posterior Predictive Distribution).

The example shows a normal mixed model (not a glimmix model).

BR, Koen