Asymptotic expansions in sequential estimation for the first-order random coefficient autoregressive model: Regenerative approach (Q1323532)

The direct purpose of this paper are Edgeworth expansions for the distribution of a sequential least squares estimator considered in the framework of a first order ergodic random coefficient autoregression (RCA). Its indirect purpose is to illustrate once again that the regenerative approach is quite suitable for the problem of second order asymptotic analysis in Markov type statistical models. The proof of the main results of this paper consists of the following two steps. First, the author places the RCA process in an appropriate Markov chain context and applies the splitting technique of \textit{K. B. Athreya} and \textit{P. Ney} [see Trans. Am. Math. Soc. 245, 493-501 (1978; Zbl 0397.60053); see also Ann. Probab. 6, 788-797 (1978; Zbl 0397.60052)] and \textit{E. Nummelin} [Z. Wahrscheinlichkeitstheorie Verw. Geb. 43, 309-318 (1978; Zbl 0364.60104)] to extend the underlying Markov chain to a regenerative sequence. This makes possible to apply asymptotic expansions within the framework of randomly stopped sequences with record regeneration. Secondly, the author goes back to the original RCA model and rewrites the expansions, obtained in the first step and expressed in terms of the regeneration cycles, in terms of the initial RCA. In this step he uses the interior structure of the RCA model and such technical devices as Wald's identities for Markov chains.