A new composite bootstrap approach for estimating the long-memory parameter

This study investigates whether combining different bootstrap methods can enhance performance, focusing on long-memory time series. Specifically, we examine a preliminary Monte Carlo experiment integrating various established para- metric and non-parametric bootstrap approaches. By linearly combining the outputs of these methods, we aim to develop a novel combined bootstrap technique called composite bootstrap. First results show improved accuracy and reliability, opening the door to methodological advancements.