LIKELIHOOD-BASED INFERENCE FOR SPATIALLY CORRELATED ORDERED CATEGORICAL DATA MODELS USING THE SAEM ALGORITHM

Spatially correlated ordered categorical data are produced by a variety of research areas, such as ecology, epidemiology and the social sciences. Such data usually arising from a continuous distribution that is clipped via a set of predetermined threshold values. This paper deals with determining maximum likelihood estimation of parameters in clipped Gaussian spatial models. For this purpose, a stochastic approximation EM algorithm is described. Finally, the predictive distribution at unsample sites is approximated based on MCMC samples.