ASYMMETRIC ANALYSIS OF DATA BY STATISTICA L IMPLICATION

Many methods of data analysis built data organisations according to a criterion of resemblance measured by different indices. Mostly these i ndices are symmetrical. However, in multiple real situations, we need to structure a set of variables or a set of variable classes according inclusion or inference relation such that:''If a then b''. Our approa ch, inspired by the works of I.C.Lerman, gave us a direction to get an oriented classification of these variables, figured on a graph. Thus we obtain an oriented classification on different classes of variables represented by a hierarchy. In a next step, we examine the significan t nodes of the latter as well as the contribution of subjects-objects considered individually or through a categorisation.