A method of assessing the relationship between two groups of variables (for example, the relationship between three measures of a worker's ability and four measures of his or her performance). The method, suggested by Hotelling in 1936, identifies the two linear combinations of variables (one from each group) that have the maximum correlation.