REVIEWING THE STATISTICAL MEASURES OF ASSOCIATION FOR THE TRIPLE DICHOTOMY.

This article reviews the statistical measures of association for the triple dichotomy. In three-variate analysis, the most interesting case occurs when the relation between variables A and B is controlled for a third one variable C, or antecedent variable. A and B are the original variables; a coefficient which is followed by AB is a zero-order coefficient; a coefficient which is followed by AB. C is a first-order coefficient referred to the AB relation when variable C is controlled. Statistics have for a long time ignored the distinction between symmetrical and asymmetrical relations. This has not produced serious consequences for bivariate relations except if an inferential interpretation was sought. The problem can be solved by paying greater attention to the hypothesized relations, and choosing the coefficient accordingly. After all, the choice is very simple: a relation can be symmetrical or asymmetrical. The range of possible relations between three variables is much greater: first, all relations are symmetric. Second, all relations are asymmetric; problems arise as regards which of the variables is the dependent one in the three bivariate relations. The possible combinations are eight. Third, some relations are symmetrical and some are asymmetrical. The possible combinations are eighteen.