Article révisé par les pairs
Résumé : Two mode analysis of a two array table makes it possible to process and to represent the lines and the columns in the same way, while factor analysis only gives a representation of the variables. The latter has the advantage, however, that by ignoring the specificities of the variables, it reduces the number of dimensions of the representation space while keeping a sufficient approximation. It is suggested that both advantages be combined by performing two mode analysis after replacing each variable by its image (in Guttman's sense) in the set formed by the other ones. Taking as an example a questionnaire about workers' styles, three methods are compared: factor analysis, two mode analysis following partitioning of the population into groups through cluster analysis of raw data, two mode analysis following cluster analysis applied to the table of images. The last method is superior. It gives two dimensions for workers' style, which it calls 'involvement in workers' activity' and 'involvement in own trade'.