DOI: 10.1140/epjst/e2008-00833-5
Selection of recurrence threshold for signal detection
S. Schinkel1, O. Dimigen2 and N. Marwan1, 31 Interdisciplinary Centre for Dynamics of Complex Systems, University of Potsdam, 14415 Potsdam, Germany
2 Department of Psychology, University of Potsdam, 14415 Potsdam, Germany
3 Potsdam Institute for Climate Impact Research (PIK), 14412 Potsdam, Germany
schinkel@agnld.uni-potsdam.de
Abstract
Over the last years recurrence plots (RPs) and recurrence quantification analysis
(RQA) have become quite popular in various branches of science. One key problem in applying RPs and RQA
is the selection of suitable parameters for the data under investigation. Whereas various well-established
methods for the selection of embedding parameters exists, the question of choosing an appropriate
threshold has not yet been answered satisfactorily. The recommendations found in the literature are
rather rules of thumb than actual guidelines. In this paper we address the issue of threshold selection
in RP/RQA. The core criterion for choosing a threshold is the power in signal detection that
threshold yields. We will validate our approach by applying it to model as well as real-life data.
© EDP Sciences, Springer-Verlag 2008