人民大学高级计量经济学讲义12.ppt

Econometric Theory Lecturer: Dr.Jingtao Yi Room 715 Business School, RUC Lecture 12: k-Variable Linear Regression X 1. Heteroscedasticity; 2. Tests for Heteroscedasticity; 3. Estimation under Heteroscedasticity. Heteroscedasticity Heteroscedasticity Properties of OLS Estimators Heteroscedasticity Estimation of var(b) Heteroscedasticity Estimation of var(b) Heteroscedasticity Estimation of var(b) Tests for Heteroscedasticity Heteroscedasticity causes inefficiency of the OLS. Four major tests: The White test The Breusch-Pagan/Godfrey test The Goldfeld-Quandt test The LR test for grouped data Tests for Heteroscedasticity The White Test Approach: Computes an auxiliary regression of the squared OLS residuals on a constant and all variables of the regressors, their squares and their cross products. Tests for Heteroscedasticity The White Test Tests for Heteroscedasticity The Breusch-Pagan/Godfrey Test Tests for Heteroscedasticity The Breusch-Pagan/Godfrey Test Tests for Heteroscedasticity The Goldfeld-Quandt Test Tests for Heteroscedasticity The LR Test for grouped data Estimation under Heteroscedasticity Two Procedures 1. OLS with White-adjusted standard errors; 2. GLS with knowledge of the structural form of heteroscedasticity. Estimation under Heteroscedasticity Method I. OLS with White-adjusted standard errors Approach: unbiased and consistent estimator, less efficient 1. Estimate the equation by OLS but compute the White covariance matrix; 2. Permit Wald tests of linear restrictions. Method II. GLS Approach: efficient estimator but the knowledge of the structural form of heteroscedasticity may not be available. 1. Estimate the functional form of heteroscedasticity; 2. Compute the feasible GLS estimator in an attempt to capture the efficiency of GLS. * *