A functional Hungarian construction for the sequential empirical process (Q2577025)

Starting with a short review on approximation results for empirical processes, the authors introduce the notation of a sequential empirical process and define a discretized Kiefer-Müller process, both indexed by functions \(f\) from a Hölder class \(\mathcal{H}\). For both processes a so-called Komlós-Major-Tusnády (KMT) result is stated, that is, the authors give an exponential upper bound for the supremum of the probability of the deviation between suitable versions of the sequential empirical process and the discretized Kiefer-Müller process. The supremum is taken (outside the probability) with respect to \(f\in\mathcal{H}\). Compared to the coupling in sup-norm, this allows for larger functional classes \(\mathcal{H}\). The result has an application in the theory of statistical experiments, where asymptotic equivalence is understood in the sense of Le Cam.