On possible dependence structures of a set of random variables (Q452848)

Let \(I\) be a countable set, let \(\mathcal{A}\subset 2^{I},\) and let \(f_{i},\) \(i\in I,\) probability densities on \(\mathbb{R}.\) Using a perturbation method, the authors construct a collection \(X_{i},\) \(i\in I,\) of random variables possessing the following properties: (i) for each \(A\in \mathcal{A},\) the random variables \(X_{j},\) \(j\in A,\) are independent, and the distribution of \(X_{j}\) has density \(f_{j},\) \(j\in A\); and (ii) for any \(B\in \mathcal{A}^{c}\) and \(C\subset B,\) the random variables \(X_{l},\) \(l\in C,\) are not independent.