论文部分内容阅读
本文建立一个状态数目由数据决定的马尔可夫转换向量自回归模型,用贝叶斯方法推断模型参数,并利用基于Gibbs分块采样的MCMC方法做逼近。然后本文用此模型和估计方法分析上海A股市场周收益率,结果发现,我国股票市场最可能存在5个不同的状态,状态间的区分首以波动性大小不同为标准,股市除了在初期波动性极小外,从1992年4月开始可以分为两个阶段,在各阶段股市均在三个状态之间转换。
In this paper, a state-dependent Markov transformation vector autoregressive model is established. The Bayesian method is used to infer the model parameters and the MCMC method based on Gibbs block sampling is used to make the approximation. Then this paper uses this model and the estimation method to analyze the weekly A-share market rate of return in Shanghai, and finds that there are 5 different states in the stock market in our country. The difference between the first and the second is based on the different volatility. From the very beginning of April 1992, it can be divided into two stages, in which the stock market changes between the three states.