Mariano Matilla-Garcia, Universidad Nacional de Educacion a Distancia (UNED), Madrid, Spain
Rina B. Ojeda, Universidad Aut noma de Cohauila, Mexico
Manuel Ruiz Marin, Universidad Politecnica de Cartagena, Madrid, Spain

Abstract: This paper suggests new nonparametric statistical tools and procedures for modeling linear and nonlinear univariate economic and financial processes. In particular, the tools presented help in selecting relevant lags in the model description of a general linear or nonlinear time series; that is, nonlinear models are not a restriction. The tests seem to be robust to the selection of free parameters. We also show that the test can be used as a diagnostic tool for well-defined models.

Keywords: lag, selection, nonlinearity, model selection, correlation integral, BDS test