Based on detecting 'shocks' (local environment changes) and tracing their effects to determine causal direction.

University of Michigan - Ann Arbor

Multiple methods exist that can be used to determine causal relationships between variables.

Causal Discovery

Pearl, Judea. “The Seven Tools of Causal Inference, with Reflections on Machine Learning.” Communications of the ACM, vol. 62, no. 3, 2019, pp. 54–60., doi:10.1145/3241036.

https://dl.acm.org/doi/pdf/10.1145/3241036

The Seven Tools of Causal Inference, with Reﬂections on Machine Learning

Provides connection between causes and probabilities. Helps enumerate testable implications of a causal model i.e can allow it to be represented in a compact set which allows for queries to be made into that specific set and thus more realistic estimates.

d-separation criterion

Used to infer causal direction by modeling P(y) as a convolution of two non Gaussian distributions, in a linear model X -> Y with non Gaussian noise. P(x) is arguably less Gaussian than P(y), and this 'less' can be measured numerically to give evidence of a direction of causality.

Learn Before

Related