[position in research by date]
[position in research by subject]
"On the Robust Estimation of Econometric
Models," Annals of
Economic and Social Measurement, October 1974, 667-677.
pdf file (673KB).
Abstract
The computational aspects of obtaining robust estimates of a general
nonlinear econometric model are described, and some results of
estimating a particular model are presented. When robust estimators are
considered as weighted-least-squares estimators, it clearly appears
feasible, by a combination of solving unconstrained optimization
problems and iterating, to obtain robust estimates of econometric
models. In estimating the particular model, the robust
estimators performed well in terms of prediction accuracy.
Comments
This paper discusses how robust estimates of econometric models can
be obtained. For the example my
forecasting model--1971#5--was estimated by
six extimators, four of them robust estimators, and the estimators were
compared in terms of predictive accuracy.
The discussion of the computation of the LAD and 2SLAD
estimators in Chapter 6,
Section 6.5.4, in 1984#2 is based on this paper.
The discussion of the LAD and 2SLAD estimators
in Chapter 6, Sections 6.3.5 and 6.3.6,
in 1984#2 is an
extension of the discussion in this paper. The 2SLAD estimator in
Section 6.3.6 for q = 1.0 is the estimator suggested at the bottom
of page 670 of this paper.
The LAD and 2SLAD estimators are programmed into the Fair-Parke
program. The commands are simply LAD and 2SLAD.