Adaptive Nonparametric Distribution-free Procedures In Factorial Data Analysis

Abstract

Many statisticians have questioned the basic assumptions about underlying models which might dominate the analysis of the data in many cases. The assumption of normality without much thought is of concern to a growing group of statisticians. If wrongly assumed, the assumption of normality can lead in serious flaws in the analysis of data. It therefore becomes important to consider distribution-free procedures that don't have to rely on the normality assumption. This is where the adaptive procedures come into play. When data is skewed or light tailed, these adaptive methods produce better results than the regular Wilcoxon and parametric methods. The problem has been solved for a c-sample problem (Sun 1997). Our goal here is to extend this method, to the two-way Anova problem.