A law of the iterated logarithm for Markov chains on \(\mathbb{N}_ 0\) associated with orthogonal polynomials (Q1322497)

The starting point are limit laws (LLN, CLT, LIL) obtained by the author for polynomial hypergroups and more general, for random walks connected with orthogonal polynomials [ibid. 3, No. 2, 245-266 (1990; Zbl 0719.60010), Adv. Appl. Probab. 22, No. 1, 68-87 (1990; Zbl 0719.60009), J. Multivariate Anal. 34, No. 2, 290-322 (1990; Zbl 0722.60021) and Monatsh. Math. 109, No. 4, 311-326 (1990; Zbl 0713.60011)]. In this paper the author considers Markov chains on the nonnegative integers. If the transition probabilities fulfil a three term recursion, a convolution structure is defined via the corresponding orthogonal polynomials. A priori no positivity is assumed, so the convolution need not to define a hypergroup. So the limit laws proved in this paper have a wide field of applications. The main theorems stated as Theorem 3.1 (for polynomial hypergroups), Theorem 3.2 (for a general class of random walks (r.w.) defined by a convolution structure) and Theorem 3.3 (for certain birth and death r.w.) are versions of the classical LIL, where the normalizations depend on the underlying structure. The formulations are to technical to be written explicitly in a review. The proofs are given in \S 4, containing a version of the CLT (Theorem 4.1) with estimation of the rate of convergence for polynomial hypergroups, which is of independent interest.