Partial transfer entropy on rank vectors

摘要: For the evaluation of information flow in bivariate time series, measures have been employed, such as transfer entropy (TE), symbolic (STE), defined similarly to TE but on ranks components reconstructed vectors, and rank vectors (TERV), similar STE forming for future samples response system with regard current vector. Here we extend TERV multivariate account presence confounding variables, called partial (PTERV). We investigate asymptotic properties PTERV, also (PSTE), construct parametric significance tests under approximations Gaussian gamma null distributions, show that cannot achieve power randomization test using time-shifted surrogates. Using simulations known coupled dynamical systems applying tests, PTERV performs better than PSTE worse (PTE). However, unlike PTE, is robust drifts series it not affected by level detrending.