MODIFIED DOMINANCE ANALYSIS

El-Houssainy Abd El-Bar Rady; Ahmed Amin ELSheikh; Mohamed Rashed Ezzeldin

Author(s)

El-Houssainy Abd El-Bar Rady Department of Applied Statistics and Econometric, Faculty of Graduate Studies for Statistical Research, Cairo University,Egyp

Ahmed Amin ELSheikh Department of Statistics, Deanship of Educational Services,Qassim University, Saudi Arabia.

Mohamed Rashed Ezzeldin Department of Statistics, Deanship of Educational Services,Qassim University, Saudi Arabia.

Abstract

Relative importance analysis is a very useful supplement to regression analysis. The purpose of determining predictor importance is not model selection but rather uncovering the individual contributions of predictors relative to each other within a selected model. Dominance analysis considers one of the most powerful relative important techniques which provides a framework for assessing the relative importance of predictors in univariate and multivariate multiple regression models. This method is based on pairwise comparisons of the contribution of predictors in all possible subset models. One of the great weaknesses of Dominance analysis (DA) that is, it will become more computationally difficult due to the exponentially growing number of sub-models involved that is due to the fact for predictors, there are (2??-1) sub-models. Here we try to extend the current studies practice by developing a technique to overcomes this disadvantage. We determined that the new technique gives the same result for classical general dominance and furthermore, we get a huge reduction in the sub-models, for the result to be robust a simulation study has been conducted.

Publication Details

Page(s) 363-378

DOI DOI not available

Published Journal: Pakistan Journal of Statistics, Volume: 36, Issue: 4, Year: 2020

Request the full text from PASTIC

Keywords

multiple regression, Predictor Importance., Relative importance, Statistical significance Simulation, Dominance Analysis