Study On The Heteroscedasticity In Linear Regression Model With Cross-Sectional Data

Regression analysis of the problems in the social economy often using cross-section data analysis is called cross section data regression analysis.As a result of the cross section data itself is easy to get the final regression model into heteroscedasticity re-gression model,and homoscedasticity in classical regression analysis model is a ba-sic assumption,therefore study how to determine whether there is in the process of cross section data regression analysis heteroscedasticity and correct heteroscedastici-ty is very important and meaningful.If there is homoscedasticity in the regression model,through the ordinary least squares parameter estimation can be linear unbi-ased estimator effectively,which can simplify the whole step regression analysis;Once heteroscedasticity regression model of out without being noticed and then continue to use ordinary least squares estimate its good least squares will be damage leading to incorrect inferences.Therefore,the inspection and correction for heteroscedasticity for further study,discriminant covered by phenomenon of authenticity,and promote the model test of simulation,as it is necessary to improve the accuracy of the regression model of step.Studying heteroscedasticity testing method and the method of variance method have great significance.The commonly used method for solving the heteroscedasticity problem is the gen-eralized least squares method when the cross section data tested heteroscedasticity,the model of the correction,effect plays an important effect on the choice of weight,now commonly used several kinds of weighting is based on the simulation model of residual values and explain the relationship between the variable structure,due to the effect of correction of the lever coefficient,so this article is presented based on the leverage coef-ficient and replace the new weight structure form of thought.In this article,through a large number of numerical simulation and example analysis proves that the weighting structure form is effective and practical.This paper firstly introduces the heteroscedasticity in the process of cross section data and expounds the possible serious consequences of directly using the ordinary least square method to deal with the heteroscedasticity model of several data surfaces.The methods of heteroscedasticity test and correction in cross section data regression analysis are reviewed.The test methods include graphical test and analytical test.Analytical test introduces the Spearman rank correlation coefficient test,G-Q test,park test,Glejser test,B-P test,white test and its generalization of K-B test.This article main content is a new structure of the weight in weighted least-square method which based on the Leverage and the ideas of Park test,inspects the effect different heteroscedasticity testing method and heteroscedasticity correction method by Monte carlo simulation experiment and the three cases.Especially under different weight through the weighted least square method to the correction effect of heteroscedasticity.The comparative analysis results show that this kind of weight construction is more simple and saves a lot of time in experiments,and the calculation amount is smaller than the traditional weight construction and the correction effect is also better.Finally,the summary and thinking of heteroscedasticity are briefly introduced.