Journal of the Japan Statistical Society, Japanese Issue
Online ISSN : 2189-1478
Print ISSN : 0389-5602
ISSN-L : 0389-5602
ON SOME R-AND M-ESTIMATORS OF REGRESSION PARAMETERS UNDER UNCERTAIN RESTRICTION
A. K. Md. Ehsanes SalehTaka-aki Shiraishi
Author information
JOURNAL FREE ACCESS

1989 Volume 19 Issue 2 Pages 129-137

Details
Abstract
For a univariate linear model Yn=Xnθ+εn, when it is apriori suspected that =h may hold, we are interested in robust estimates of θ. Four classes of point estimations, namely, the unrestricted R-estimators (URE), restricted R-estimators (RRE), the preliminary test R-estimators (PTRE) and shrinkage R-estimators (SRE) are considered. In the light of asymptotic distributional risks, their relative efficiency results under local alternatives are studied in details. Although the SRE may dominate the URE, neither of the SRE nor PTRE dominate each other. We may conclude that the SRE and PTRE are robust relative to the URE and RRE. Furtheremore four versions of M-estimatiors corresponding to the URE, RRE, PTRE and SRE are proposed and the results similar to the cases of the R-estimators are discussed briefly.
Content from these authors
© Japan Statistical Society
Previous article Next article
feedback
Top