topic Re: How to output 200K parameters in PROC MCMC? in Statistical Procedures

How to output 200K parameters in PROC MCMC?

AndreaB_ — Wed, 02 Jan 2019 17:12:14 GMT

Hi,

I want to use PROC MCMC to run a random-effects linear negative binomial (NB1) model and estimate ~200,000 random intercepts (as an alternative to NLMIXED). Since the OUTPOST= dataset can only save up to 32,767 variables, what approaches can I take to be able to output the posterior samples for all of the ~200K random-effects parameters?

Thanks,

Andrea

P.S. I have SAS v9.4 (TS1M4) and SAS/STAT 14.2.

Re: How to output 200K parameters in PROC MCMC?

Garnett — Wed, 16 Jan 2019 00:26:46 GMT

Hi,

An easy way is to assume the posterior on each effect is Gaussian:

1) Monitor the posterior distributions of the random effects.

2) Write ODS output postsumint=Px;

3) In a subsequent datastep, sample from the mean and Standard deviation given in Px (or just use the posterior info in Px.)

Let me know if this isn't what you are thinking...