Mean, median or something else

Abstract

Non parametric inferential statistics was used to guide the numerical optimization study to search for the optimum result in this article. The technique was demonstrated in a case study and a significant improved predictive model was formulated with 4% less residual sum of squares (RSS) than the median model, 20% less than the mean model and 79% less than a benchmarked empirical model. The methodology proposed herewith addressed the selection dilemma between mean and median. It identified an optimum value and formulated a better predictive model than those by either mean or median.