If, as above, an association appears to exist, other variables may be introduced by further dividing the categories (e.g. by sex and age). The purpose of displaying data in this way is not only to examine what relationship there may be between variables, but to explore this relationship by the further breakdowns that may be made, e.g. by eliminating 'S purious’ relationships, explained by the prior influence of a third variable, or 'S pecifying’ the interaction between three or more variables. See also LAZARSFELD, CORRELATION, MULTIVARIATE ANALYSIS.