A Nonparametric Causality Test: Detection of Direct Causal Effects in Multivariate Systems Using Corrected Partial Transfer Entropy

摘要: In a recent work we proposed the corrected transfer entropy (CTE), which reduces bias in estimation of (TE), measure Granger causality for bivariate time series making use conditional mutual information. An extension TE to account presence other is partial (PTE). Here, propose correction PTE, termed Corrected PTE (CPTE), similar way CTE: shifted surrogates are used order quantify and correct bias, involved entropies high-dimensional variables made with method k-nearest neighbors. CPTE evaluated on coupled stochastic systems both linear nonlinear interactions. Finally, apply economic data investigate whether can detect direct causal effects among variables.