On the strong law of large numbers for functionals of countable nonhomogeneous Markov chains (Q1327558)

A strong law of large numbers for functionals of nonhomogeneous Markov chains is proved. The approach is analytic and different from the usual one. The main idea is to construct a suitable monotone function and to use Legendre's theorem on a.s. differentiability of monotone functions. We formulate one of the results: Let \(\{f_ n(X_ n), n\geq 0\}\) be a functional of a nonhomogeneous countable Markov chain \(\{X_ n\}\) and \(0<a_ n\uparrow\infty\). Let \(\{g_ n, n\geq 0\}\) be a sequence of positive, even and continuous functions on \(R\) such that, as \(| x|\) increases, \(g_ n(x)/| x|\uparrow\), \(g_ n(x)/x^ 2\downarrow\), \(n\geq 0\). If \(\sum_ n E(g_ n(f_ n(X_ n)))/g_ n(a_ n)<\infty\), then \(\sum_ n {f_ n(X_ n)-E(f_ n(X_ n) \mid X_{n-k})\over a_ n}\) converges a.s., and \({1\over a}_ n\sum^ n_{m=1}(f_ m(X_ m)-E(f_ m(X_ m)\mid X_{m-k}))\to 0\) a.s. for all \(k\geq 1\) (where \(X_{-n}=\text{const.}\), \(n\geq 1\)).