广义自回归条件异方差模型的异常点检验OACHSSCD

This paper studies the detection of outliers in a generalized autoregressive conditional heteroscedasticity model un-der normal distribution.Based on the mean shift model and variance weighted model,the expression of the score test statistic and its asymptotic distribution are given,and the effectiveness of this test method is confirmed through numerical simulation.In empirical study,the daily data of the New York Stock Exchange Composite Index are chosen to construct a generalized autoregressive condi-tional heteroscedasticity model,and the differences between the score test method and the local influence analysis method in the di-agnosis of this model are compared and analyzed.