The central limit theorem for Tukey's 3R smoother (Q1186027)

In his discussion of exploratory analysis of time series \textit{J. W. Tukey} [Exploratory data analysis (1977; Zbl 0409.62003)] defined and advocated a nonlinear smoothing algorithm in which the running median of three consecutive quantities is applied repeatedly until no further change occurs. The algorithm, called the 3R smoother, has the properties of being resistant to occasional outliers and at the same time responsive to changes in trend. Several authors discussed some aspects of the 3R smoother for sequences of weak dependent random variables. The aim of this paper is to analyse asymptotic properties of the nonlinear smoothing algorithm. It is shown that under appropriate conditions the output of the algorithm is absolutely regular and that the partial sums are asymptotically normal.