Bayesian Prediction in Clipped GLG Random Field Using Slice Sampling

Journal of Statistical Theory and Applications (JSTA)

关键词: Binary spatial data;    Bayesian latent model;    Heavy tail;    Slice sampling;   

DOI  :  10.2991/jsta.2014.13.2.5

来源: DOAJ

【 摘 要 】

By assuming that an underlying Gaussian-Log Gaussian (GLG) random field clipped to yield binary spatial data, we propose a new model which provides flexibility in capturing the effects of heavy tail in latent variables. For our analysis, we adopt a Bayesian framework and develop a Markov chain Monte Carlo (MCMC) algorithm to carry out the posterior computations. Specifically, we introduce auxiliary variables and employ the slice sampling method to simulate from the full conditional distribution of components which does not define a standard probability distribution. Then, the predictive distribution at unsampled sites is approximated based on acquired samples. Finally, we illustrate our methodology considering simulation and real data sets.

【 授权许可】

Unknown