lect6_103_2010_Compatibility_Mode_b

# lect6_103_2010_Compatibility_Mode_b - 5 Additional Topics...

This preview shows pages 1–4. Sign up to view the full content.

4/14/2010 1 5 Additional Topics in Basic Regression Analysis • 1) Heteroskedasticity/BLUE • 2) Measuring Model Fit • 3) Changing Units of Measurement • 4) Dummy Variables • 5) Omitted Variable Bias Heteroskedasticity/BLUE - I • Our regression model is: • Recall that u i represents variables other than X i that affect Y i . u i is a random variable, just like X i and Y i . Let’s think about its variance. • More specifically, let’s think about its conditional variance, i.e. 0 1 = i i i Y X u β + + ( 29 i i Var u X c =

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
4/14/2010 2 Heteroskedasticity/BLUE – II • What exactly is the conditional variance: • It tells us the what the variance of u i is given that X i equals some number c, i.e. – Var( u i | X i = 5) is the variance of u i if X i = 5. – Var( u i | X i = 20) is the variance of u i if X i = 20. • Sometimes Var( u i | X i =c) is the same for every possible value of X i (i.e. for every possible c) Other times, Var( u i | X i =c) varies depending on the particular value of X i (i.e c). ( 29 i i Var u X c = Heteroskedasticity/BLUE - III • Defn: • 1) If Var( u i | X i =c) is the same for every value c (i.e. the variance of u i does not depend on X i ), then u i is homoskedastic . • 2) If Var( u i | X i =c) does depend on c, then u i is heteroskedastic . • Example: #WebsiteHits = β 0 + β 1 Advertising\$ + u • There are two important implications of this distinction between homoskedasticity and heteroskedasticity.
4/14/2010 3 Heteroskedasticity/BLUE - IV • If u i is homoskedastic, then: • 1) The OLS estimator β 1 is BLUE, i.e. it is the B est, L inear, conditionally U nbiased E stimator (of β 1 ). – “Best” means that the OLS estimator β 1 is the most efficient of these estimators, i.e. the one with the smallest variance. • 2) We can actually use a slightly simpler formula for Var( β 1 ) and SE( β 1 ) (again, you don’t need to remember this exact formula) ( 29 ( 29 ( 29 ( 29 2 1 1 1 1 2 1 1 ˆ 1 2 ˆ ˆ ˆ Var SE Var 1 n i i n i i u n n X X n β = = - = = - Heteroskedasticity/BLUE - V • If u i is heteroskedastic, then: • 1) The OLS estimator β 1 is not BLUE, i.e. there are potentially more efficient other estimators out there (specifically, something called weighted least squares – but we will not cover that in this class)

This preview has intentionally blurred sections. Sign up to view the full version.

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

{[ snackBarMessage ]}