5 Growing Libraries in Python for Causality Analysis

At least in neuroscience, mainly 2 models have been mostly used (or abused): Granger causality and Dynamic causal modeling (DCM). Those are, at the same time, the Good, the Bad, and the Ugly. Granger causality (so called from its inventor) is based on the idea that if one time series (called the "predictor") has some predictive power for another time series (called the "predictand"), then past values of the predictor should contain information that can help forecast future values of the predictand. To test for Granger causality, researchers typically use statistical techniques such as regression analysis or vector autoregression (VAR) to estimate the strength of the relationship between the two-time series. If the predictor series is found to significantly improve the forecast of the predictand series, it is said to have Granger causality with respect to the predictand.