Lecture 7 heteroscedasticity emet20076007 24 th april

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ticity funtion as weights in WLS ) bi = exp b0 + b1 x1 + b2 x2 + h α δ δ + bk xk δ Feasible GLS is consistent and asymptotically more e¢ cient than OLS. Lecture 7 (heteroscedasticity) EMET2007/6007 24 th April 2013 25 / 34 Example: Demand for cigarettes Estimation by OLS d cigs = 3.64 + 0.880 log (income ) (24.08 ) 0.501 educ (0.167 ) n = 807 0.751 log (cigprice ) (0.728 ) 0.771 age (0.160 ) (5.773 ) 0.009 age 2 (0.0017 ) 2.83 restaurant (1.11 ) R 2 = 0.0526 cigs Cigarettes smoked per day income Income cigprice Cigarette price restaurant Smoking restrictions in restaurants p valueBreusch Pagan = 0.00 Reject homoscedasticity Lecture 7 (heteroscedasticity) EMET2007/6007 24 th April 2013 26 / 34 Estimation by FGLS d cigs = 5.64 + 1.30 log (income ) (17.80 ) 0.463 educ + 0.482 age (0.120 ) n = 807 income 2.94 log (cigprice ) (0.44 ) (0.097 ) (4.46 ) 0.0056age 2 (0.0009 ) 3.46 restaurant (0.80 ) R 2 = 0.1134 Now statistically signi…cant Discussion: The income elasticity is now statistically signi…cant; other coe¢ cients are also more precisely estimated (without changing qualitative results except for age) Lecture 7 (heteroscedasticity) EMET2007/6007 24 th April 2013 27 / 34 What if the assumed heteroscedasticity function is wrong? If the heteroscedasticity function is misspeci…ed, WLS is still consistent under MLR.1 – MLR.4, but robust standard errors should be computed Recall the following assumptions: Assumption MLR.4 (Zero conditional mean) E (εi jx1,i , x2,i , . . . , xk ,i ) = E (...
View Full Document

This note was uploaded on 06/15/2013 for the course EMET 2007 taught by Professor Strachan during the Two '13 term at Australian National University.

Ask a homework question - tutors are online