期刊名称：Brazilian Journal of Probability and Statistics
印刷版ISSN：0103-0752
出版年度：2006
卷号：20
期号：1
页码：67-91
出版社：Brazilian Statistical Association
摘要：We examine the class of extended generalized inverse Gaus-sian (EGIG) distributions. This class of distributions, which appeared brie.yin a monograph by J.rgensen (1982), is more deeply and broadly studied inthis paper. We start by deriving its probabilistic properties. Furthermore,we use the EGIG family in two popular and important statistical problems,one from survival analysis and the other from financial econometrics. In thefirst one, a survival model, we compare our results with those obtained byAchcar and Bolfarine (1986) and found out that the EGIG explains better thedata at the cost of one extra parameter. In the second case, we extend thetraditional univariate stochastic volatility model by allowing the errors of thesquared returns of the BOVESPA index to be driven by a EGIG distribution.We compared our results with those found by Lopes and Migon (2002) andfind once again that the EGIG is able to identify and estimate the quantitiesusually searched by standard univariate sto chastic volatility mo dels, such ashighly persistent and skewed log-volatilities. Markov chain Monte Carlo arespecifically tailored for both applications and a variant of the slice sampler isused to sample from some of the full conditionals
