ESTIMATION OF THE PARAMETERS OF A BINARY MARKOV RANDOM-FIELD ON A GRAPH WITH APPLICATION TO FIBER-TYPE DISTRIBUTIONS IN A MUSCLE CROSS-SECTION

Methods are discussed for the estimation of the parameters of a binary Markov random field (BMRF) defined on a graph. The standard method is maximum pseudo-likelihood (MPL) estimation. Maximum likelihood (ML) e stimation has been hampered in the past by the intractability of the l ikelihood function. Recently Markov chain Monte Carlo (MCMC) methods h ave been introduced for ML estimation. In this paper a new method for Monte Carlo maximum likelihood is described. It is used for the estima tion of the parameters of a simple model (the Ising model of statistic al physics). As an application the distribution of fibre types in a cr oss-section of human muscle is analysed.