The following features have been added to the COUNTREG procedure:
Bayesian estimation features
possibility of choosing the prior distributions through the PRIOR statement
several tools to control and optimize the initialization and the tuning phase
multithreaded Metropolis sampling
convergence diagnostic tools: Raftery-Lewis, Heidelberger-Welch, Geweke, effective sample size
an automated reiterative implementation of the MCMC algorithm that can search for the stationary region of the posterior distribution
an automated reiterative implementation of the MCMC algorithm that can determine the length of the posterior samples
a weighted scheme that can recombine the samples from multiple MCMC chains when each chain approximates only a portion of the posterior distribution
A PERFORMANCE statement has been added, enabling you to control the number of threads that are used during optimization.
The item store now supports BY groups. Parameter estimates and other results can be stored in an item store on a per-BY-group basis and later retrieved for scoring the BY groups in another data set.