New asymptotic results for generalized Oppenheim expansions (Q6097389)

The paper under review deals with the convergence issue for weighted partial sums of random variables related to generalized Oppenheim expansions from probability and almost surely perspectives. Particular cases of Lüroth series, Engel series and Sylvester series have been studied before by several authors. Let \(B_n\) be a sequence of integer valued random variables defined on \((\Omega =[0, 1], \sigma_{[0,1]} = {\mathcal{A}}, P)\) for a Lebesgue measure on \([0, 1]\), and \(\phi_n: \mathbb N^* \to \mathbb R^+\) be a sequence of functions with \(q_n =q_n(h_1, \ldots, h_n)\) be a sequence of nonnegative numbers such that \(h_1\ge 1, h_j \ge \phi_j (h_{j-1})\) for \(j=2, \ldots, n\). Define \[ R_n = \frac{1}{\delta_n (B_n, B_{n+1}, Q_n)} = \frac{B_{n+1} + \phi_n (B_n) Q_n}{\phi_n (B_n)(1+Q_n)}. \] The purpose of this paper is the study of the asymptotic behavior of \(\frac{1}{b_n^p} \sum_{j=1}^n a_j (R_j - Eg (R_j))\) for a suitable chosen constant \(p, a_n, b_n\) and function \(g (\cdot)\). This follows the same line of \textit{T. Nakata} [Bull. Inst. Math., Acad. Sin. (N. S.) 15, No. 2, 123--141 (2020; Zbl 1448.60098)] and \textit{R. Giuliano} [Monatsh. Math. 187, No. 3, 509--530 (2018; Zbl 1410.11108)] who established the weak law for the sequence \(\frac{1}{n \log n} \sum_{j=1}^n R_j\) and the authors [J. Theor. Probab. 34, No. 3, 1579--1606 (2021; Zbl 1472.60056)] for exact weak and strong laws. Section 2 prepares some preliminary results from [Zbl 1472.60056] to get lower and upper bounds for the probability \(P(R_n > x)\) in Lemma 2.1, the upper bound for the expectation \(E(R_n^q I(R_n\le t))\) in Lemma 2.2 via integration by parts, Lemma 2.3 shows the identity \(\lim_{n\to \infty} \sum_{j=1}^n P (R_j > \frac{b_n}{a_j}) =0\) holds for certain conditions on the distribution and sequences \(a_n\) and \(b_n\), and Lemma 2.5 for the upper bound of \(E(R_{nj}^p)\) for \(R_{nj} = R_jI(R_j\le \frac{b_n}{a_j}) + \frac{b_n}{a_j} I(R_j > \frac{b_n}{a_j}) \) for both case \(p>\alpha\) and \(p=\alpha\) by using previous lemmas. Section 3 is devoted to the weak law of large numbers, i.e., the convergence in probability of the sequence \(\frac{1}{b_n^p} \sum_{j=1}^n a_j (R_j - Eg (R_{nj}))\) in Theorem 3.1 under the conditions of Lemma 2.3 for \(a_n\) and \(b_n\) and conditions on distributions \(F_j\) through the decomposition \(R_j - Eg (R_{nj}) = (R_j - R_{nj}) + (R_{nj} - Eg(R_{nj}))\) starting with Markov inequality, previous lemmas and earlier work of the authors. Theorem 3.3 and Theorem 3.6 for the same result hold with \(\alpha=1\) and \(p=1\), respectively, again by decomposing terms. As applications of Theorem 3.1, Theorem 3.8 extends to triangular array of real numbers with special properties with \(\sum_{j=1}^{m_n}c_{nj} (R_j - E(R_jI(\vert c_{nj}R_j\vert \le 1))) \to 0\) as \(n\to \infty\), and Theorem 3.10 reduces to earlier results. Section 4 provides a strong law of large numbers for the sequence \(R_n\) in Theorem 4.1: \(\frac{1}{\rho(n) \log^{\beta} n} \sum_{j=1}^n \frac{\log^{\beta - p}j}{j}R_j \to 0\) a.s. when \(n\to \infty\) for \(\beta>0\) and \(p\ge 2\) and every sequence \(\rho (n)\) with the convergence of the series t\(\sum_{n=1}^{\infty}1/\rho^2 (n)\). The proof of this strong law can also be found in Theorem 11 of [Zbl 1472.60056] but for different sequences of positive numbers. It would be interesting to see if the techniques in the paper can be extended for all semi-heavy tailed random variables.