On Characterizing the Multivariate Linear Exponential Distribution1

Abstract

If x and y are independent p component column vectors, and the conditional distribution of x, given x+y = z, is known, what can be said about the distributions of x and y? This problem has been solved by Seshadri (1966) in the particular case when the conditional distribution of x, given x+y = z, is multivariate normal. In fact Seshadri′s paper implicitly contains a characterization of the multivariate linear exponential distribution(1)where A(x) is a function of x not involving the p component column vector w of constant terms.