No label defined (Q3973916)

[For the entire collection see Zbl 0722.00032.] Based on the observation of the first \(n\) random variables \(X_ 1,\ldots,X_ n\) from a strictly stationary, discrete-time Markov process \(X\) the author investigates the asymptotic properties of a natural nonparametric estimate of the one-step transition probability distribution function as \(n\) tends to infinity. (The estimate is obtained by integrating the ratio of kernel estimates for the stationary two- and one-dimensional distributions.) Under appropriate conditions it is shown that this estimate is uniformly (in the main argument) strongly consistent and asymptotically normal; further, a natural estimate for the p-th quantile \((p\in(0,1))\) is shown to be strongly consistent. Besides certain regularity assumptions a key assumption is that \(X\) has only one ergodic set, which has no cyclically moving subsets, and fulfils a \(\rho\)-mixing condition where \(\rho\) decreases of polynomal order. [Editorial remark: See also the author's article reviewed below.].