Abstract:

This document describes a new method of selecting variables based on the data theory. It appears, indeed, that every day we are faced with prolific data some of which, individually, or in combinations, could be ignored without detriment to the concepts which it is expected to see ensuing from the data as a whole. Moreover, the method attempts to take into consideration the aggregate data contained in the data as a whole, and to select a minimum of sub-divisional variables summarising that data in the best possible way. The uses of this are numerous, and we propose to discuss certain of these. We describe, in this document, the method, as well as an application purposely chosen in a sphere other other that of market surveys : indeed, we have processed, by means of the RESUM programme, certain medical data, which underwent, moreover, a correspondance analysis.