It is shown by a simple argument that all recurrence relations and identities satisfied by distribution functions of order statistics arising from exchangeable random variables also hold good for averages of such distribution functions for arbitrary random variables.

Starting from any univariate density function f(x) and an associated orthogonal function g(x), we present in this note a method of constructing multivariate distributions of all dimensions. It is shown that these multivariate distributions, in addition to being closed under marginal and...