基于AR和TAR模型的变点问题分析

Analysis of change-point problem with AR and TAR models

在正态分布下,Perreault等刻画了水文时间序列4种情形的变点问题,但没有利用到具体的模型.本文设定自回归(AR)和门限自回归(TAR)模型对应均值的变化,利用贝叶斯推断法,完成这4种情形下变点问题的分析.模拟实验的结果比较理想.

Under the hypothesis of normal distribution,Perreault et al depicted the change-point problems of hydrometeorological time series in four cases according to mean and variance changing.However,it is simple and has not used any concrete model.Basing upon it,we join the AR and TAR models in the mean-equations to finish the analysis of four types of change-point problems using Bayesian inference method.Simulation experiments indicate that our method can achieve ideal results.