ASYMPTOTIC-BEHAVIOR OF THE GIBBS SAMPLER

被引:47
|
作者
CHAN, KS
机构
关键词
BAYESIAN ANALYSIS; CONJUGATE PRIOR; EXPONENTIAL FAMILY; GEOMETRIC ERGODICITY; INCOMPLETE DATA; MARKOV CHAIN;
D O I
10.2307/2290727
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
Mild sufficient conditions for the ergodicity of the Gibbs sampler are obtained. The geometric ergodicity of the Gibbs sampler is then discussed, where the main tool is the ''drift'' criterion. This criterion requires the construction of a generalized energy function so that the chain is dissipating energy, at a suitable rate, whenever it lies outside the ''center'' of the state-space. For the ''two-variable'' case, a useful recipe is provided for constructing appropriate energy functions. These results are illustrated with four examples.
引用
收藏
页码:320 / 326
页数:7
相关论文
共 50 条