Abstract
A method to analyse a l x k matrix (z = || zij ||) is introduced. The method “weighted binary analysis” has been developed in the theoretical framework of Faverge’s classical binary analysis and is based on a weighted expression of distances (instead of the classical Euclidian distance) So it makes possible to take into account differences in importance, which must be given to each row and/or column in the determination of proximities or distances (as it may be the ease if we arc more or less confident in data or if they do not have a same importance according to a given aim). Two examples of utilization of the method are given. The first example is a fictitious example which clearly shows the effect of weights on the representation (Tab. 1 and Fig. 1). The second example is in fact a proposal of application of the method to questionnaires on attitudes and opinions. The proposal is to collect next to the score matrix z a confidence matrix c (subjects’ confidences in their answers to the questions), and then to use the first dimension of the binary analysis of matrix c as weights which will affect the spatial representation of the rows and columns of z. An empirical example is giver (Annexe and Fig. 2, 3, 4, and 5).