Data Analysis And Robust Statistical Inference

Date of Award

1982

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Abstract

The performance of three cross-validatory estimators is compared with that of 11 well-known robust estimators considered by Stigler {lcub}1977{rcub}, by application to 24 real data sets.;In multiple regression, each observation is tested to see if it is an outlier. Cross-validatory procedure is then proposed to assign weights to those observations detected as outliers for the estimation of parameters.;The robustness of four predictors including a proposed new one which does not involve covariances of the order statistics, based on censored samples from the Weibull distribution is examined.;For the estimation of the centre of a symmetric distribution which is unknown, two robust estimators are proposed and their performance is investigated.;In the estimation of the three parameter Weibull distribution function it appears that almost similar distributions can be obtained by using a different set of estimators. This possibility is further examined by using graphical techniques.;In the one parameter distribution a test procedure based on prediction is proposed to detect outliers and the distribution of the test statistic is obtained.

This document is currently not available here.

Share

COinS