Research on Nonparametric Methods for Linear Models.

Abstract

The goal of this project was to conduct research in the development of nonparametric statistical methods based on rank statistics for the general linear model. Consideration was to be given to the use of a comprehensive framework to handle estimation, confidence regions and tests of hypotheses for a wide range of multiple regression and analysis of variance problems. Methods were sought which would be efficient and convenient to use in practice. The practical aspects of implementation for applied problems requires special attention in order that the methodology be direct in interpretation and computationally feasible.