Study On Bayesian Method Of Markov Switching Vector Autoregression Model

Markov regime switching model is an important nonlinear model,which is always used to describe data in financial markets.This paper used markov switching vector autoregression model to describe the data structure of financial time series.In the process of estimating the parameters of the model,we adopt the bayesian method to estimate the parameters in order to make up for the inadequacy of the traditional maximum likelihood estimation method.We got the simulation results of the estimate by MCMC method and Gibbs sampling,and then through the empirical analysis of the relationship between RMB exchange rate and Shanghai and Shenzhen stock market,we verified the rationality and validity of the method.The paper is divided into five parts:1.The paper introduces the markov regime switching models,the bayesian theory,MCMC method and some related testing methods,which lays a theoretical foundation for the formal research of this paper.2.By analyzing the structure of markov switching vector autoregression model,the paper set the prior probability of the parameter and use the bayesian method to infer the posterior distribution of parameters,and then using Gibbs sampling to estimate the model parameters.3.In the empirical analysis,the paper constructe the markov switching vector autoregression model between the exchange rate of RMB against the dollar and the return of Shanghai and Shenzhen stock market.By setting the prior distribution of model parameters,we use bayesian method for parameter estimation,and get posterior densities of the parameters.By analyzing the estimated results,we conclude the dynamic relationship between the RMB exchange rate and the stock market.Finally,we diagnose the convergence of markov chains obtained by MCMC method,and analysis whether the results are valid.