Symbolic Correlation Integral. Getting Rid of the Proximity Parameter (Draft)
Mardi | 2014-02-04 salle des thèses M Victoria CABALLERO – Mariano MATILLA GARCIA – Manuel RUIZ-MARIN In this paper we introduce the symbolic correlation integral SC(m), which avoids the noisy parameter ε of the classical correlation integral defined by Grassberger-Procaccia. Moreover we provide the asymptotic distribution of SC(m) under the null of i.i.d.. With a MonteCarlo simulation we show the size and the power performance of the new test under linear and nonlinear processes.