When does \(E(X^{k}\cdot Y^{l})=E(X^{k})\cdot E(Y^{l})\) imply independence? (Q2495418)

The authors study various conditions when the moment equality \(E( X^k Y^l) = E (X^k) E(Y^l)\) for all \(k,l \in \mathbb{N}\) for a bivariate random variable \((X,Y)\) implies independence of \(X\) and \(Y\). In particular, let \(K\) and \(L\) be subsets of \(\mathbb{R}\), and suppose that \(X\) and \(Y\) take only values in \(K\) and \(L\), respectively, and suppose further that the distributions of \(X\) and \(Y\) are determined by their integer moments \(E (X^k)\) (\(k\in\mathbb{N}\)) and \(E(Y^k)\) (\(k\in\mathbb{N}\)), respectively. Then it is shown that the equation above implies independence of \(X\) and \(Y\). A converse of this theorem is also obtained, showing that whenever \(X\) is not determined by its integer moments and the moments of \(Y\) do not correspond to those of a discrete measure, then the equation above does not imply independence of \(X\) and \(Y\).