A1313
Title: Causality with or without the Reichenbach principle: The degrees of freedom method
Authors: Abstract: The purpose is to introduce a novel methodology for uncovering causal interactions between deterministic or stochastic dynamic systems, with a particular focus on distinguishing true causal relationships from misleading correlations arising due to hidden common causes. The proposed approach is applicable to both discrete and continuous systems in time and space. Central to the method is random forest clustering of coupled 0-1 Markov chains, through which causal relationships between systems are classified directly from observed data. The effectiveness of the technique is demonstrated on both simulated and real-world time series. Notably, the method reveals a new result: the number of sunspots influences global Earth temperature with zero lag on a monthly scale. To the best of knowledge, prior studies have reported minimal lags of two to three years, with some suggesting even longer intervals.
