基于M估计强混合重尾序列结构变点的鲁棒检验OA北大核心CHSSCDCSSCICSTPCD

In view of the detection of change points of strongly mixed heavy tail sequence structures,and in order to avoid the ordinary least square deviation caused by a heavy-tailed sequence,this paper proposes an M-estimate-based ratio-type statistic to test the change point with a heavy-tailed sequence.Under general constraints,it is proved that the limit distribution of the sta-tistics under the null hypothesis is the functionality of Brownian motion,and the consistency under the alternative hypothesis is ob-tained.Aiming at the experience level distortion caused by sequence dependence,the Block Bootstrap sampling method is used to obtain a more accurate critical value,which effectively improves the inspection efficiency.The numerical simulation results show that the M-estimation-based ratio-type test under the Block Bootstrap sampling method can better control the experience level and reasonable experience potential in the change point detection of strongly mixed heavy-tailed sequence structure.Finally,the feasibility of the proposed test method is verified by a set of exchange rate data.