Note on the Unbiasedness of a Mixed Regression Estimator

THE PRESENT note is a sequel to earlier articles by Suzuki [2] and Nagar and Kakwani [1] which, among other things, deal with the bias of a member of the f class of mixed regression estimates proposed by Theil [3]. Suzuki [2] proved that if the regression disturbances are normally distributed the bi...

Full description

Saved in:
Bibliographic Details
Published in:Econometrica 1968-07, Vol.36 (3/4), p.610-611
Main Author: Kakwani, N. C.
Format: Article
Language:English
Subjects:
Bias
Estimation bias
Estimators
Matrices
Notes and Comments
Unbiased estimators
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:THE PRESENT note is a sequel to earlier articles by Suzuki [2] and Nagar and Kakwani [1] which, among other things, deal with the bias of a member of the f class of mixed regression estimates proposed by Theil [3]. Suzuki [2] proved that if the regression disturbances are normally distributed the bias of the member of the f class (for which f=1/s2, s2 being the un-biased estimator of 2, the residual variance) is of smaller order than 1/T, T being the number of observations. Nagar and Kakwani [1] derived the bias of this member of the f class without the assumption of the normality of disturbances and showed that under the normality (or even symmetry) assumption, the bias to order 1/T is zero.
ISSN:0012-9682
1468-0262
DOI:10.2307/1909528