Asymptotic Behaviour of the Error Probabilities in the Pseudo-Likelihood Ratio Test for Gibbs-Markov Distributions
Martin Janžura · Contributions to statistics · 1994
For testing simple statistical hypotheses given by Gibbs-Markov random fields a Bayes test based on the pseudo-likelihood ratio is constructed in order to deal with numerically available quantities. The asymptotic behaviour of the error probabilities is investigated with the aid of an appropriate version of the large deviations theorem. Under natural assumptions the error probabilities are proved to tend to zero exponentially fast.