Unlike the most widely used methods of researching equivalent vertices
in network analysis (structural or regular equivalences), the one whi
ch we propose in.this paper is based upon a new definition of equivale
nce meeting the following requirements: (1) analyzing the data from a
double point of view: local and global; (2) conserving the specificity
of discrete data. Applying our algorithm in the context of a sociabil
ity survey allows us to pinpoint groups with sociological coherence. T
he study shows that network differences correspond to socio-demographi
c differences.