統計学輪講（第２７回）

日時      2009年11月17日(火)    15時～15時50分
場所      経済学部新棟３階第３教室
講演者    Victor De Oliveira (The University of Texas)
演題      Bayesian Default Analysis for Gaussian Random Fields

概要

Gaussian random fields are useful mathematical tools for modeling spatially varying
phenomena and are often the default model of choice (possibly after a transformation).
The Bayesian approach for the analysis of geostatistical data has become more common
in recent years, but specification of the prior distribution for these models is a
somewhat challenging task. On the one hand, it is difficult to carry out subjective
elicitation of the prior distribution, either because of lack of prior information
or the difficulty in interpreting some of the parameters, and on the other hand
naive specification of the prior distribution may give rise to improper posterior
distributions. This talk provides a review of the main results obtained in the
last decade on objective (default) Bayesian methods for the analysis of spatial data
using Gaussian random fields.