This is a strong point from Jonathan Foulds PhD 👇🏽 The main reason these studies are done or published is to imply causation. In that sense, they are a negative scientific effort: the generation of spurious false-positives. #TypeIerror
Model-Free Sequential Testing for Conditional Independence via Testing by Betting deepai.org/publication/mo… by Shalev Shaer et al. including Yaniv Romano #TypeIerror #MachineLearning
From Shapley back to Pearson: Hypothesis Testing via the Shapley Value deepai.org/publication/fr… by Jacopo Teneggi et al. including Yaniv Romano #TypeIerror #NeuralNetwork
Learning to Increase the Power of Conditional Randomization Tests deepai.org/publication/le… by Shalev Shaer and Yaniv Romano #TypeIerror #MachineLearning