02-04-2016 03:20 PM
How to decide number of Markov's Chain iterations while running proc MCMC for bayesian approach of multiple linear regression?
In my datasset it seems that I am unable to remove autocorrelation from the model. Please suggest.
02-05-2016 11:52 AM
Your convergence might be slow. Consider increasing your burn-in iterations and thining your chain by a factor of, let's say, 10 or 20.
PROC MCMC DATA = YourDATA nbi=50000 nmc=200000 thin =20 ;
Your posterior estimates will be based on 200,000/20 = 10,000 observations.