This preview shows pages 1–9. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Econometrics 1 Lecture 4: The CLR model with normally distributed errors CLR model with normal errors We consider the CLR model + = X y with the additional assumption ) , ( ~ 2 I N i.e. the nvector of random errors has a multivariate normal distribution with mean 0 and variance matrix I 2 . Econometrics 2 Sampling distributions of b and 2 s . Because ' ) ' ( 1 X X X b + = we have ( ) 1 2 ) ' ( , ~  X X N X b Econometrics 3 Next, because = M e e 2 ' we have by a result for a quadratic function in a normal random vector ) ( ~  ' 2 2 K n X e e For the joint distribution of b and 2 s , it can be shown that b and 2 s are independent. Econometrics 4 Conclusion: ( ) 1 2 ) ' ( , ~  X X N X b ) ( ~  ) ( 2 2 2 K n X s K n b and 2 s are independent (given X ) In general if ) 1 , ( ~ N z ) ( ~ 2 K v v z , stochastically independent then ) ( ~ K t K v z t = has a (Student) tdistribution with K degrees of freedom. Econometrics 5 Because K is the only parameter the distribution can be tabulated. Econometrics 6 Econometrics 7 Let k b be the k th regression coefficient with sampling variance 1 2 ) ' ( kk X X with 1 ) ' ( kk X X the k th diagonal element of 1 ) ' ( X X . Hence ) 1 , ( ~  ) ' ( 1 N X X X b kk k k and ) ( ~ ) ( 2 2 2 K n s K n Because these random variables are stochastically independent, we have that given X the ratio ) ( ~ ) ' ( ) ' ( 1 2 2 1 K n t X X s b s X X b kk k k kk k k = Econometrics 8 Because the tdistribution does not depend on X , the result is also true unconditionally....
View Full
Document
 Fall '07
 Rashidian
 Econometrics

Click to edit the document details