Complete moment convergence and \(L_q\) convergence for AANA random variables under sub-linear expectations (Q6178667)

Given a measurable space \((\Omega, \mathcal{F})\), consider a collection \(\mathcal{P}\) of probability measures on \((\Omega, \mathcal{F})\). Now, define the following upper and lower probabilities \[ \mathbb{P}(A) = \sup_{P \in \mathcal{P}} P(A), \text{ and } \mathsf{P}(A) = \inf_{P \in \mathcal{P}} P(A), \tag{1} \] for all \(A\in \mathcal{F}\). Upper and lower expectations of a random variable \(X\) are defined analogously, \[ \mathbb{E}(X) = \sup_{P \in \mathcal{P}} E_P(X), \text{ and } \mathsf{E}(X) = \inf_{P \in \mathcal{P}} E_P(X),\tag{2} \] where \(E_P\) denotes standard expectation with respect to a probability measure \(P\). The Choquet expectations for a random variable \(X\) are defined as \begin{align*} C_{\mathbb{P}}(X) &= \int_{0}^\infty \mathbb{P}(X \ge t) \mathrm{d}t + \int_{-\infty}^0 ( \mathbb{P}(X \ge t) -1)\mathrm{d}t, \\ C_{\mathsf{P}}(X) &= \int_{0}^\infty \mathsf{P}(X \ge t) \mathrm{d}t + \int_{-\infty}^0 ( \mathsf{P}(X \ge t) -1)\mathrm{d}t. \end{align*} The primary contribution in this paper is a limit theorem for row-wise asymptotically almost negatively associated (AANA) random variables. A collection of random variables \(\{X_n, n \ge 1\}\) is called AANA under \(\mathbb{E}\) if there exists a sequence of dominating coefficients \(\{u_n, n\ge 1\}\) such that \(u_n \to 0\) as \(n\to \infty\) and if the following holds \begin{align*} \mathbb{E}(f(X_n) g(X_{n+1}, X_{n+2}, \ldots, X_{n+k})) &\le u_n \sqrt{\mathbb{E}( f(X_n) - \mathbb{E}(f(X_n)) )^2} \\ &\quad \times \sqrt{\mathbb{E}( g(X_{n+1}, X_{n+2}, \ldots, X_{n+k}) - \mathbb{E}(g(X_{n+1}, X_{n+2}, \ldots, X_{n+k})) )^2 } \end{align*} for all coordinate-wise non-decreasing or non-increasing functions \(f\), and \(g\). An array \(\{X_{n, k}, n\ge 1, k\ge 1\}\) of random variables is called row-wise AANA random variables if the sequence of random variables \(\{X_{n, k}, k \ge 1\}\) are AANA for every \(n\ge 1\). The authors prove that under certain moment assumptions about a mean-zero row-wise AANA random variables \(\{X_{n, k}, n\ge 1, k\ge 1\}\) that \[ C_{\mathbb{P}} (a_n^{-1} \max_{1 \le j \le n} |{\sum_{k=1}^{j} X_{n, k} }| )^{q} \to 0 \] as \(n\to \infty\), for a sequence of positive numbers \(\{a_n, n\ge 1\}\) and \(q\ge 1\). Results of this form have gained some attention in the recent years, primarily as a means to model uncertainty about the probability measures.