MCMC Algorithms for constrained variance matrices

Authors
Browne, W.J.
Year
2006
Journal
Computational Statistics and Data Analysis, 50:7, 1655-1677
DOI
10.1016/j.csda.2005.02.008
Abstract

The problem of finding a generic algorithm for applying Markov chain Monte Carlo (MCMC) estimation procedures to statistical models that include variance matrices with additional parameter constraints is considered. Such problems can be split between additional constraints across variance matrices and within variance matrices. The case of additional constraints across variance matrices is considered here for the first time and a review of existing work on the case of additional parameter constraints within a variance matrix is given. Two simple single-site updating random walk Metropolis algorithms are described which have the advantage of generality in that they can be applied to virtually all scenarios. Four applications where these methods can be used in practice are given. Some situations when such single-site algorithms break down are described and multiple-site alternatives are briefly discussed.

Number of levels
2
Model data structure
Response types
Multivariate response model?
Yes
Longitudinal data?
Yes
Further model keywords
Substantive discipline
Substantive keywords
Impact

Paper describes MCMC algorithms for problems with constrained variance matrices

Paper submitted by
William Browne, Bristol Veterinary School, University of Bristol, william.browne@bristol.ac.uk
Edit this page