L'analyse des variables indépendantes continues et catégorielles : alternatives à la dichotomisation

Abstract

Summary : The analysis of continuous and categorical independent variables : Alternatives to dichotomization. Transforming continuous independent variables into categorial ones makes the statistical analyses simpler because after dichotomization, the effects of these variables can be examined via an analysis of variance (ANOVA) rather than a multiple regression analysis. However, this simplicity comes at a high price. When a continuous variable is dichotomized, one artificially introduces random error which decreases the statistical power of the inferential analyses. Assuming a normal distribution, the decrease in statistical power is equivalent to the exclusion of approximately 38 % of the participants. In this article, we present the problems associated with the dichotomization of continuous variables and we discuss various strategies that allow researchers to analyze experimental designs with continuous and categorical independent variables. Key words : continuous variables, quantitative variables, regression analysis, mean deviation form, dichotomization.