Detecting Causality from Nonlinear Dynamics with Short-term Time Series

摘要: Quantifying causality between variables from observed time series data is of great importance in various disciplines but also a challenging task, especially when the are short. Unlike conventional methods, we find it possible to detect only with very short data, based on embedding theory an attractor for nonlinear dynamics. Specifically, first show that measuring smoothness cross map two can be used causal relation. Then, provide effective algorithm computationally evaluate map, or “Cross Map Smoothness” (CMS) and thus infer causality, which achieve high accuracy even data. Analysis both mathematical models benchmarks real biological systems validates our method.