Nodal Analysis of Associated Populations

A METHOD of subdividing an associated population the individuals of which are specified by the presence or absence of a number of attributes has been described previously1; the process is ultimately derived from the R-techniques of factor analysis, and the groups of individuals so obtained will be called R-groups. The attributes may similarly be subdivided into Q-groups by the corresponding Q-techniques. There results a 2-way table in which every attribute-individual record is referred to one of each kind of group. The requirement is to extract from this table the most concentrated groups of records, the attribute-individual noda around which the population may be regarded as varying. The problem is essentially that of establishing coincidences between R and Q factors; even in normal factor analysis the coincidences are impaired in the formation of the correlation matrix2,3, and in a subdivision process involving successive correlation matrices the divergence may well be considerable.