Meesum: I wouldn't call what I can tell you an "expert opinion". You'll have to take it for what it's worth.

In the example below I specified in PROC MI the number of MCMC imputations (replicates) with the Nimpute setting. The source dataset, ExperimentU, included some rows with a missing value for the DV. PROC MI imputes missing DV values as well as missing IV values in the five (in this case) MCMC replicates. I then used the imputed dataset, ExperimentUI, and PROC REG to estimate model parameters, and PROC MIANALYZE to adjust estimated variances.

proc mi data=ExperimentU seed=337131 Nimpute=5 round=.01 minmaxiter=20 out=ExperimentUI ; MCMC outest=MCMC /*ACFPlot=WLF*/ ; EM maxiter= 100 /* ; var predM x1 x2 x3; run; quit;

ods trace on; proc reg data=ExperimentUI outest=parmests covout noprint; model predM = x1 x2 x3/ covb ; by _imputation_; output p=ests; ods output covb=covbests; run; quit;

proc mianalyze data=parmests; modeleffects Intercept x1 x2 x3; ; run; ods trace off;

S

-----Original Message----- From: owner-sas-l@listserv.uga.edu [mailto:owner-sas-l@listserv.uga.edu] On Behalf Of meesum kazmi Sent: Tuesday, April 15, 2008 5:45 AM To: SAS-L@listserv.uga.edu Subject: Help Needed in simualting the MCMC

Hi All, Need your help in Simulating the MCMC model. Just to give the background i tried runnning multiple linear regression on different time frame data, as a result of this excercise i got coeffecients for the independent variables. now to get a robust estimates of these variables i want to simulate the MCMC on the these coeffecients. Need your help in validating the methodology, and wants expert opinion on how can we simulate this in SAS. There is experimental proc in SAS 9.2 which simulates MCMC and we can also use proc mi but currently i am not sure on how to use these procedures. Any help would be highly appreciated

Best, Meesum