This article discusses the differences between predictive and causal inference, explains why correlation does not imply causation, and why machine learning is not inherently suited for causal inference. It highlights the limitations of using machine learning for causal estimation and provides suggestions for when each type of inference should be used. The article also touches on causal machine learning and its role in addressing the challenges of high-dimensional data and complex functional forms.
We measured both undirected FC (correlation in the time domain, coherence in the frequency domain) and directed FC (Granger causality, in both time and frequency domains) on the same data.
