Measurement And Influence Analysis Of Leverage Point In Linear Model With AR(1) Error

Regression diagnosis is a very important issue in statistical research.In recent years,the study of leverage measurement and influence analysis in regression diagnosis has attracted much attention.This paper mainly studies the leverage measurement and influence analysis of several estimators in linear model with AR(1)error and complex collinearity.The specific contents are as follows:Firstly,the influence measurement matrix of ridge estimator,principal component estimator and r-k estimator in linear model with AR(1)error is mainly presented,and the leverage measurement and related properties of principal component estimator and r-k estimator are explored.The leverage effect of the first sample point in the regression space by principal component estimator,r-k estimator.Finally,the change of leverage measurement with the change of related parameters is obtained.Secondly,the modified r-d estimator is proposed by combining the prior information with r-d estimator,and the influence measurement matrix of the modified r-k estimator and the modified r-d estimator is defined by using the method of quasi-projection matrix.At the same time,through example and simulation analysis,this paper studied the leverage measurement and influence of the modified r-k estimator and the modified r-d estimator,and discussed the leverage effect of the first sample point in the regression space when the modified r-k estimator and the modified r-d estimator were used.Finally,the influence of different parameters on the modified r-k estimator and the modified r-d estimator leverage measurement is explored,and the size of each estimator leverage measurement value is compared.