期刊名称：Brazilian Journal of Probability and Statistics
印刷版ISSN：0103-0752
出版年度：2014
卷号：28
期号：2
页码：190-208
DOI：10.1214/12-BJPS201
语种：English
出版社：Brazilian Statistical Association
摘要：We propose a dynamic model to analyze polychotomous data subject to temporal variation. In particular, we propose to model categorized levels of rainfall across time. Our model assumes that the observed category is related to an underlying latent continuous variable, which is modelled according to a power transformation of a Gaussian latent process, centered on a predictor that assigns dynamic effects to observable covariates. The inference procedure is based on the Bayesian paradigm and makes use of Markov chain Monte Carlo methods. We analyze artificial sets of data and daily measurements of rainfall in Rio de Janeiro, Brazil. When compared to the fitting of the actual observed volume of rainfall, our categorized model seems to recover well the structure of the data.
