Concentration Ellipsoids, Their Planes of Support, and the Linear Regression Model

Abstract

The relationship between the concentration ellipsoid of a random vector and its planes of support is exploited to provide a geometric derivation and interpretation of existing results for a general form of the linear regression model. In particular, the planes of support whose points of tangency to the ellipsoid are contained in the range (or column space) of the design matrix are the source of all linear unbiased minimum variance estimators. The connection between this idea and estimators based on projections is explored, as is also its use in obtaining and interpreting some existing relative efficiency results.

Keywords:

• BLU estimators
• Concentration ellipsoid
• Linear regression
• Planes of support

JEL Classification:

• C1
• C2
• C3

ACKNOWLEDGMENT

I would like to thank the editor and a referee for useful comments and sugggestions.