This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 241B Lecture Generalized Least Squares One classic assumption (1.4) states that the errors have a spherical conditional covariance matrix, E ( UU j X ) = & 2 I: Today we relax the assumption and allow E ( UU j X ) = & 2 V ( X ) ; (1) where V ( X ) is an n & n nonsingular matrix. Under (1) the errors are (con ditionally) heteroskedastic and the elements of V ( X ) are, in general, nonlinear functions of X . Di/ering diagonal elements correspond to heteroskedasticity and nonzero o/diagonal elements correspond to correlation. Consequence of Relaxing Classic Assumption 1.4 There are 3 main consequences for the OLS estimator The GaussMarkov Theorem no longer holds, hence the BLUE estimator is not the OLS estimator. The OLS estimator remains unbiased. The t test statistic no longer has a t distribution. (Same for the F test statistic.) Homoskedasticity: Conditional versus Unconditional There are three "levels" of homoskedasticity. The simplest case is one in which the error is conditionally homoskedastic...
View Full
Document
 Fall '08
 Staff

Click to edit the document details