Abstract
Given the sequence of variables x0,x1,…,xm we show how to test the hypothesis that, for each j and t < j, the conditional distribution of xj given (x0,x1,…,xt) is free of (x0,x1,…,xt−1). Our assumptions are that the population from which the sequence is drawn has a multivariate normal density and that the population may be sampled repeatedly.