In statistics, a procedure for optimal classification of individuals into groups or classes on the basis of a number of discriminating variables on which each of the individuals has been measured. For example, a researcher might use the discriminating variables age, sex, level of education, and income in order to discriminate optimally between people who support capital punishment and those who oppose it. This is achieved by weighting the variables and combining them into discriminant functions that separate the two groups maximally. This procedure is mathematically equivalent to canonical correlation with the discriminating variables considered as predictor variables and the group membership variable considered as a dummy criterion variable taking on arbitrary values (0 and 1 in the case of two groups). Alternatively, it may be thought of as multiple regression with a categorical dependent variable. Also called discriminant analysis. DFA abbrev.