基于计算密集型方法的动态广义线性模型研究

Statistical inference in dynamic generalized linear models, has recently become a very hot topic. Here the focus of the study, is the inference of computation-intensive methods for dynamic generalized linear models. Based on the previous research work, the proposed methods are the computation-intensive methods, in order to improve the accuracy and timeliness of statistical computation. We study the corresponding theory and real data analysis of statistical computing.. In the statistical computing theory, we present several basic conditions and some basic principles, solving dynamic generalized linear models by the computation-intensive methods; in the economic and financial related empirical analysis, we introduce the basic standards of parameter coefficient selection for the computation-intensive methods. This research will not only theoretically clarify the basis of computation-intensive methods for dynamic generalized linear models, but also provide the new ways of data analysis for the models.

动态广义线性模型的统计推断，已成为近来研究的热点问题。这里重点研究，动态广义线性模型数据的计算密集型解法问题。为了提高统计计算的效率，拟采用计算密集型方法解该模型，并进行相应的理论和数据分析研究。对于计算密集型解法的统计计算理论，重点给出计算密集型解法如（并行MCMC），解动态广义线性模型的基本原则；在数据分析上，给出计算密集型方法的基本标准。本研究不仅能从理论上阐明动态广义线性模型计算密集型解法的基础，而且能够为该模型的数据分析提供新途径。

对计算密集型方法，特别是，并行MCMC 和MCMC方法，处理动态广义线性模型进行研究，尤其对该类方法的统计计算理论和统计计算效率；同时提高了对动态广义线性模型兼容性的理论认识。重点对泊松型动态广义线性模型的经济和金融应用进行研究。由于金融数据具有时间相依性，有些情况满足该模型的要求，故对金融数据应用研究是必要且合理的。