The goal of this project is to present a classical and Bayesian study in spatialmodels for censored data using more robust distributions than the normal andskew-normal distribution, i.e., using the scale mixture of skew-normal class of distributions. Furthermore, it will be present classical and Bayesian diagnostic studies based in local influence methods (Cook, 1986) and the q-divergence (Peng & Dey, 1995), respectively, as discuss in Lachos et al. (2011) and Lachos et al. (2013). For the estimation step, we will use EM (Expectation-Maximization), SAEM (Stochastic Approximation of the EM) and the Gibbs Sampler algorithms with implementation in R, C++ and WinBugs.The proposal of this project looks for contributing positively to the developing ofspatial models for censored data, providing new results in models of practical interest, extending and complementing some previous results found in Militino and Ugarte (1999); Kim and Mallick (2004);De Oliveira (2005);Fridley and Dixon (2006); Toscas (2010); Karimi and Mohammadzadeh (2012); Prates et al. (2012); Assumpção et al. (2014); Prates et al. (2014); Schelin and Sjostedt-de Luna (2014); De Bastianiet al. (2014), among others.
News published in Agência FAPESP Newsletter about the scholarship: