Research On Heteroscedastic Data With Change Points

At present,scholars at home and abroad study a lot of heteroscedasticity or change points,but there are few studies on heteroscedastic data with variable points,and the latest research on heteroscedasticity is mainly used to study which statistics are used to measure heteroscedasticity models.The variance of the parameters is estimated to be better at some stage.Research on how to better handle data with heteroscedastic effects is not detailed enough.Moreover,for the heteroscedastic data with the change point,the change point and the heteroscedastic interaction and mutual influence,if only one of them is considered,the estimated model difference is also great.Therefore,this paper studies the processing method of heteroscedastic data with variable points,and proposes a new HCCMEs-based model transformation method to process the original heteroscedastic data.This method is combined with the change point detection method to process the heteroscedastic data with the change point.For the heteroscedastic data with variable points,this paper explores which index has more influence on the model,that is to say,using the variable point detection method and the HCCMEs-based model transformation method to detect the heteroscedastic data with the change point,through the detection The model goodness of fit is used to judge the heteroscedastic data for the change point.It is good to perform the change of the change point and then to perform the heteroscedasticity elimination effect or to perform the heteroscedasticity elimination and then the change point detection effect.By setting two different parameters β,manually set the change point,simulate the data with different parameters of the same model before and after the change point,and use the model:yi=β1+β2Xl+σiui,i=1,……,n.Using Monte Carlo simulation with data with heteroscedastic effect,the heteroscedastic data is estimated by the four HCCMEs statistics of HCO,HC2,HC3,and HC4 for the heteroscedastic error variance sequence of the whole sample data,and four heteroscedastic differences are obtained.Residual sequence.The model is transformed by the four statistical statistic pairs to obtain the heteroscedasticity to achieve the purpose of processing the heteroscedastic data,and then the binarized segmentation method is used to analyze the heteroscedastic data with the change points,deal with.After a lot of experiments,the results show that the model transformation method based on HCCMEs has a good effect on the elimination of heteroscedasticity.The degree of heteroscedasticity after the heterogeneous variance is eliminated by the model transformation is greatly reduced compared with the degree of heteroscedasticity before the heteroscedasticity is eliminated.Combined with the HCCMEs-based model transformation and binarization segmentation method,the heteroscedastic data with variable points is processed.The results show that the heteroscedastic sequence with the change point is firstly changed,and then the detected change point position The method of heterogeneous variance elimination based on HCCMEs model transformation to obtain a new model is optimal.In addition,for heteroscedastic data processing with variable points,it is found that HCCMEs have the best effect of eliminating heteroscedasticity by HC4-based model transformation,and the effect is most stable when multiple experiments are performed.