A note on the performance of a wavelet-based multiscale estimator

Abstract

Quantitative evidence is presented in order to study the performance of a wavelet-based, second order estimator of relevance to multiscaling phenomena such as telecommunications traffic. Special attention is given to the behavior of the confidence intervals of the scaling exponent under Gaussian assumptions with synthetic long-range dependent trajectories, allowing for a straightforward contrast with the theoretical results presented in the bibliography. The bias-variance tradeoff decision behind the choice of the onset of the scaling region is also revisited, along with its influence on the confidence of the intervals of the exponent.