第八章 自相关2016.ppt

Chapter 7 Heteroscedasticity This assumption is relaxed: The disturbances ui appearing in the population regression function are homoscedastic ,or they all have same variance. var(ui)=?2i=?2=constant，i=1,2,…,n heteroscedastic：the variances of disturbance vary with X, or is a function of X var(ui)= ?2 i=f(Xi) 7.1 the heteroscedasticity in real life 1、error-learning models As people learn, errors they make become smaller over time, variance of errors decrease too. 2、savings function As incomes grow ,people have more choices about their saving behavior. 3、cross-section data Cobb—Douglas product function Y=ALαKβeu AS the size become larger, the company has more choice about product techinique and management mode, the difference of product amount become larger. 7.2 the source and consequence of heteroscedasticity 1、the source of heteroscedasticity because the disturbance stands for other factors ignored by the regression model，the heteroscedasticity may result from these factors ignored. a．Specification error: some important variables are omitted from the model. the disturbance includes the influence of the important variable ignored and these variable may be related to some regressors b．Skewness in the distribution of one or more regressors included in the model. c.Measurement error and the improvement of data collecting techniques d. Presence of outlier: Outlier: an observation is much different(very small or very large) comparing with other observation. f. Cross-section data: In cross-section data, one may deal with objects with different order of magnitude 2、the consequence of heteroscedasticity a. the OLS estimator of coefficient is still linear and unbiased, but not the best,or haven’t the minimum variances。 because ?2 I vary with Xi, the variance of estimator can’t be obtained. b．It is difficult to proceed t test and establish the confidence interval of coefficient。 c． Is not unb