Density estimation in the \(L^ \infty\) norm for dependent data with applications to the Gibbs sampler (Q688378)

The author studies the behaviour of kernel estimators for a \(d\)- dimensional smooth probability density and its derivatives. The main object of this paper is to develop the rates of convergence w.r.t. the \(L^ \infty\)-norm on \([0,1]^ d\) or \(\mathbb{R}^ d\) if these estimates are applied to dependent but stationary data [see also the paper of \textit{G. G. Roussas}, J. Stat. Plann. Inference 18, No. 2, 135-149 (1988; Zbl 0658.62048) for related but different results]. In the present paper it is shown, that under certain \(\beta\)-mixing conditions the same minimax convergence rates (w.r.t. a.s.- and \(p\)- convergence) as for i.i.d. samples can be achieved. Furthermore, convergence rates in terms of the \(\beta\)-mixing conditions are presented. As an application, uniform convergence rates for problems related to the Gibbs sampler are given. One of the main tools for the proof is \textit{D. Pollard's} empirical process technique [Empirical processes: Theory and applications. (1990; Zbl 0741.60001)].