Bayesian estimate of the within and between variance with PROC MIXED ?

sbxvab · Posted 11-08-2011 06:41 AM

Dear all,

We use a mixed model approach to estimate the within and between variance.

This works fine, unless the the within variance is larger than the between variance. We do not want negative variance components (i.e. the “nobound” option in SAS proc mixed).
At the moment, we use JMP “variability charts” to obtain a Bayesian estimate of the within and between variance.

What is the equivalent proc mixed code?

Thanks a lot,

Vaélrie

SteveDenham · Posted 11-11-2011 08:22 AM

Wouldn't the REML estimates be adequate i.e., default estimates without the nobound option?

Steve Denham

Funda_SAS · Posted 12-05-2011 03:07 PM

Hi Vaélrie,

The MIXED procedure does not support the Bayesian analysis. However, you can use PROC MCMC for fitting general Bayesian models. To specify linear and nonlinear random-effects models, you need to use the new RANDOM statement in the MCMC procedure.

Below is a link to a SAS Global Forum Paper which explains how to fit random-effects models with PROC MCMC:

http://support.sas.com/resources/papers/proceedings11/334-2011.pdf

Funda