Adaptive density estimation on bounded domains under mixing conditions (Q2188474)

In this paper the classical problem is studied of the estimation of a density function \(f : D\to \mathbb{R}\), where \(D\subset \mathbb{R}^d\) from the observation of \(n\) identically distributed random variables \(X_1,\dots,X_n\) not necessarily independent. In several modern applications, \(D\) is a known bounded domain whose shape can be quite complex (e.g. geographical region). A very interesting approach is given: a specific family of kerneltype estimators are introduced to devoid boundary bias. The authors then propose a data-driven Goldenshluger and Lepski type procedure to jointly select a kernel and a bandwidth. The optimality of the given procedure in the adaptive framework, stating an oracle-type inequality is proven. The paper is organized in 7 section. A very interesting real-data example is given.