Common canonical variates

摘要: Canonical correlation analysis measures the linear relationship between two random vectors X 1 and 2 as maximum combinations of . Several generalisations canonical to k > ,...,X have been proposed in literature (Kettenring, 1971, 1985), based on principle maximising some generalised measure correlation. In this paper we propose an alternative generalisation, called common variates, assumption that variates same coefficients all sets variables. This generalisation is applicable situations where i dimension. We present normal theory likelihood estimation illustrate their use a morphometric data set.