Nonparametric Approach to Multiple Regression

Abstract

Many real life problems require the use of multiple regression to characterize and solve. Sometimes however the independent variables come in forms that violate the assumptions of the parametric multiple regression. This paper developed a nonparametric approach which uses ranks of both the dependent and independent variables to achieve the objectives of multiple regression. This approach accommodates data as low as the ordinal scale and robust. A prediction procedure which is by interpolation, is also presented.

Keywords: Rank, Robust, Prediction, Interpolation, Rank Correlation