Statistical inference, a new method of data analysis. (L’implication statistique, une nouvelle méthode d’analyse de données.) (French) Zbl 0843.62002

Summary: If the problem of the simultaneity of two variables \(a\) and \(b\) finds part of its solution in the symmetrical study of correlation or similarity, the question of inference (if \(a\) then \(b\)) has on the other hand to undergo the examining of a dissymmetrical relationship. In relation with a psychological problem of complexity, R. Gras [Thèse d’Etat, Univ. Rennes (1979)] brought a contribution that made possible a number of applications of this type of relationship in research in psychological genetics or in mathematical didactics, those areas not precluding other application fields. But the variables taken into account in his research are restricted to binary ones, presence-absence of a feature in a given individual. The purpose here is to extend the study of the statistical inference (or quasi-inference) to other types of variables and above all, to classes of such variables. This extension allows us to build a tree of oriented classes. These developments and this original construction mainly proceed from A. Larher [Thèse d’Etat, Univ. Rennes (1991)].


62-07 Data analysis (statistics) (MSC2010)
62A01 Foundations and philosophical topics in statistics
Full Text: Numdam EuDML