SLTBL模型的Bayes分析

Abstract

利用M arkov cha in M on te C arlo技术对可分离的下三角双线性模型进行B ayes分析.由于参数联合后验密度的复杂性,我们导出了所有的条件后验分布,以便利用G ibbs抽样器方法抽取后验密度的样本.特别地,由于从模型的方向向量的后验分布中直接抽样是困难的,我们特别设计了一个M etropolis-H astings算法以解决该难题.我们用仿真的方法验证了所建议方法的有效性,并成功应用于分析实际数据.

A Bayesian method is provided to analyze seperable lower-triangular bilinear time series models via Markov chain Monte Caro technique.To generate the sample by Gibbs sampler,we first derive all of the conditional posterior distributions.Since it is difficult to sample directly from the conditional posterior of the index in the models,a special Metropolis-Hastings step is designed.We demonstrate the proposed method by simulated and real examples.