Mixing limit theorems for ergodic transformations (Q2471125)

In this paper, the author firstly proves the following theorem: Let \(R_n\), \(n> 1\), be random variables on the probability space \((X,{\mathcal A}, P)\), taking values in the separable metric space \((M, d)\), and \(R\) a random element of \(M\) such that \(R_n\overset {P}\Rightarrow R\) as \(n\to\infty\). Assume there is a \(\sigma\)-finite measure \(m\) on \((X,{\mathcal A})\) such that \(P\ll m\), and an ergodic nonsingular transformation \(T\) on \((X,{\mathcal A}, m)\) under which \((R_n)_{n\geq 1}\) is asymptotically invariant in measure in that \(d(R_n\circ T, R_n)@>m>> 0\) as \(n\to\infty\). Then \(R_n\overset{{\mathcal L}(m)}\Longrightarrow R\) as \(n\to\infty\). Here, \(R_n\overset {P}\Rightarrow R\) denotes that the distributions \(P\circ R^{-1}_n\) of the \(R_n\) converges weakly to that of \(R\), and convergence in measure, \(@>m>>\), w.r.t. a \(\sigma\)-finite measure to be understood as \(\to^Q\) for all probability measure \(Q\ll m\). Furthermore, \(R_n\overset{{\mathcal L}(m)}\Longrightarrow R\) means that \(R_n\overset {Q}\Rightarrow R\) for all \(Q\ll m\). If \(m\) itself is a probability measure, this is equivalent to \(R_n\overset {m}\Rightarrow R\) (mixing) in Rényi's sense. Secondly, the author applies the theorem to obtain distributional and weak functional limit theorems for ergodic processes with arbitrary absolutely continuous initial distributions such as ergodic sums, renewal-theoretic variables and hitting times for ergodic measure preserving transformations.