[Standardization, a tool for the analysis of transversal data in epidemiology]

The direct method of standardization is mainly applied for eliminating undesirable effects of confounding variables in the comparison of several rates or means. This paper stresses the usefulness of standardization for detecting the statistical association between two variables. Under the assumption of a simple model implying two main effects and an interaction, it is discussed how the comparison between crude and standardized values measures altogether the main effect of the variable and the interaction. Furthermore, the variation in the population of the differences between crude and standardized values is shown to be an index of the effect of the variable for which it has been standardized.