Empirical quantile central limit theorems for some self-similar processes (Q2346983): Difference between revisions

Given i.i.d.\ copies \(\{X_j\}_{j=1}^{+\infty}\) of an input stochastic process \(X=\{X(t):\, t \in E\}\), the empirical distributions built on \(X\) are defined by \[ F_n(t,x) =\frac{1}{n}\sum_{i=1}^n I_{(-\infty, x]}(X_i(t)). \] Let \(F(t,x)\!:= P(X(t) \leq x)\). This paper considers empirical quantile processes \[ \left\{\sqrt{n}\biggl(F_n^{-1}(t,\alpha) - F^{-1}(t,\alpha)\biggr):\; n\geq 1\right\}, \] where \(F_n^{-1} (t,\alpha)\!:= \inf \{x:F_n(t,x) \geq \alpha\}\) and \(F^{-1} (t,\alpha)\! := \inf \{x:F(t,x) \geq \alpha\}\), and it investigates for which \(X\)'s these processes converge to a Gaussian process. This investigation was motivated by the work of \textit{J. Swanson} [Probab. Theory Relat. Fields 138, No. 1--2, 269--304 (2007; Zbl 1116.60010); Stochastic Processes Appl. 121, No. 3, 479--514 (2011; Zbl 1230.60022)], where convergence was established in the special case when the input process is a Brownian motion. In this paper, the authors extend these results by proving that, if \(X\) satisfies five given conditions, then the empirical quantile processes converge to a Gaussian process (Theorem 1). Some of these conditions deal with self-similarity and path properties of the input process, and analytic properties of \(F(t,x)\); others deal with the validity of the empirical and empirical quantile CLTs with input data determined by i.i.d.\ copies of \(X\), and bounds on the probabilities of the events \(\{\sup_{ t \in J\cap Q} |X(t)| > u\}\), where \(Q\) denotes the rational numbers. Input processes which satisfy these conditions include fractional Brownian motion and symmetric stable processes with stationary independent increments (Theorem 2). Convergence is shown to be uniform over quantiles in any closed sub-interval of \((0,1)\). In addition, the authors investigate how sample path properties of the input process influence these central limit theorems (Theorem 3). This work is a continuation of the work of the same authors in [Prog. Probab. 66, 167--194 (2013; Zbl 1271.60037)].