Generating Parallel Correlated Transition Frequencies for a Markov Chain.
Abstract
This paper describes an algorithm and a FORTRAN program called MCHAIN for simulating k parallel Monte Carlo replications of a Markov chain using rotation sampling. This method of sampling produces k sample transition frequency vectors with a desirable structure of statistical dependence among them. In particular, these sample vectors can be used to estimate the probability that, say, n sub 1, ..., n sub r transitions of types 1,...,r occur during a first passage from state a to state b with a variance of the estimate of 0(1/k squared) and a computation time 0(k) as k yields infinity. This compares favorably with the case of independent replications wherein the estimate would have a variance 0(1/k) and computation time 0(k) as k yields infinity. An example including a sample driver program are presented to illustrate how MCHAIN works in practice. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1982
- Accession Number
- ADA124250
Entities
People
- George S. Fishman
Organizations
- University of North Carolina at Chapel Hill