Construction de bases d'ondelettes de $L^20,1$ et estimation du paramètre de longue mémoire par la méthode des ondelettes.

摘要

This study is devoted to the use of wavelets in two different fields, constructions of bases on the interval and long-memory parameter estimation. In the first part we present general constructions of orthogonal and biorthogonal multiresolution analyses on the interval. In the first one, we describe a direct method to define an orthonormal multiresolution analysis. In the second one, we use the integration and derivation method for constructing a biorthogonal multiresolution analysis. As applications, we prove that these analyses are adapted to study regular functions on the interval (H^{s}([0,1])$ et $H^{s}_{0}([0,1])$ for $sinmathbb{N}$). The second part is devoted to the study of adaptive wavelet-based estimators of the long-memory parameter for Gaussian then linear processes in a general semiparametric frame. We introduce and develop the choice of a data-driven optimal bandwidth. Moreover, we establish central limit theorems for the estimators of the memory parameter with the minimax rate of convergence (up to a logarithm factor). Finally adaptive goodness-of-fit tests are also built and easy to be employed: they are chi-square type tests. Simulations confirm the interesting properties of consistency and robustness of the adaptive estimators and tests.