The authors propose a nonparametric test for white noise hypothesis in a regression model. The test is based on a regressogram estimate, which bin width is selected using a penalty method similar to the AIC, BIC, and C'ps MAllows criteria.

We propose a minimax statistical framework adapted to nonparametric estimation in deterministic dynamical system, which highlights the importance of the recurrence property of the observed process. Estimators are evaluated on the basis of a weighted uniform norm. The weight function rn depends...