On the Komlós-Révész SLLN for dependent variables (Q1714992)

Let $X_1, X_2,\dots$ be a sequence of random variables on a probability space $(\Omega,\mathcal{F},P)$ and let $\| X\|_p^p=E\vert X\vert ^p$. Kolmós and Révész (see \textit{P. Révśz} [The laws of large numbers. New York-London: Academic Press (1968; Zbl 0203.50403)]) have proved that if the $X_i$'s are independent and centered random variables, then $\frac{\sum_{k=1}^{n}\| X_k|_p^{-q}}{\sum_{k=1}^{n}\| X_k\|_p^{-q}}\longrightarrow 0$ a.s. (as $n\longrightarrow \infty)$, provided that $\sum_{k=1}^{\infty} \| X_k\|_p^{-q}$ diverges and $p=q=2$. This result has been generalized by many authors. \textit{A. Rosalsky} [Bull., Calcutta Stat. Assoc. 35, 59--66 (1986; Zbl 0611.62031)] generalized it for pairwise independent random variables and $p=2$, and \textit{G. Cohen} [Acta Sci. Math. 74, No. 3--4, 915--925 (2008; Zbl 1212.60032)] generalized it for martingale differences and $p\in (1,2]$. The author here generalizes the result for the Kolmós-Révész strong law for two dependent families of random variables. The first one is for a negatively dependent sequence, while the second dependent family is best referred to the paper itself due to the inherent technical definition. Nevertheless, the results are generalizations of the above result for two dependent families.