Discovering contemporaneous and lagged causal relations in autocorrelated nonlinear time series datasets

The paper introduces a novel conditional independence (CI) based method for and nonlinear, lagged and contemporaneous causal discovery from time series in the causally sufficient case. Existing CI-based such as the PC algorithm and also common methods from other frameworks from low recall and partially inflated false positives for strong which is an ubiquitous challenge in time series. The novel, PCMCI^+, extends PCMCI Runge et al. , 2019b to include discovery of links. PCMCI^+ improves the reliability of CI tests by the choice of conditioning sets and even benefits from. The method is order-independent and consistent in the oracle. A broad range of numerical experiments demonstrates that PCMCI^+ has adjacency detection power and especially more contemporaneous recall compared to other methods while better controlling false. Optimized conditioning sets also lead to much shorter runtimes than PC algorithm. PCMCI^+ can be of considerable use in many real world scenarios where often time resolutions are too coarse to resolve delays and strong autocorrelation is present.