基于似然比扫描方法的INGARCH模型变点探测

Change Point Detection of INGARCH Models Based on Likelihood Ratio Scanning Method

整数值时间序列数据在许多领域中非常常见,可用整数值广义自回归条件异方差(INGARCH)模型来拟合。本文研究了INGARCH模型中的变点探测问题,基于似然比扫描方法(LRSM),讨论了分段平稳的INGARCH模型中变点的数量和位置。然后通过大量数值模拟,验证LRSM的有效性,最后并将其应用于实际数据的分析中。

Integer-valued time series data are very common in many fields and can be fitted by integer-valued generalized autoregressive conditional heteroscedasticity (INGARCH) models. In this paper, the problem of change point detection in INGARCH model is studied. Based on the likelihood ratio scanning method (LRSM), the number and location of change points in piecewise stationary INGARCH process are discussed. Then the validity of LRSM is verified by a large number of numerical simulations, and finally it is applied to the analysis of real data.