Home | Archive

Bayesian networks come in equivalence classes

🌰

A set of conditional independencies can have many different Bayes nets that encode it. These together form a Markov equivalence class: i.e. a class of DAGs that share a skeleton (the undirected graph structure, \(\text{abs}(A)\) where \(A\) is the adjacency matrix) and colliders i.e. \(A \rightarrow B \leftarrow C\) where \(A\) and \(C\) are not adjacent.

In other words, you can't select a member of the class for free. The additional structure must have attached semantics of some kind.

Various regimes have been proposed. Lots of the thinking-about-agency wonkiness operationalizes to this or something bigger that contains this.

Further reading:

Footnotes:

1

Pearl (2009-09-01) Causality: Models, Reasoning and Inference, Cambridge University Press, .

2

Pearl & Mackenzie (2018-05-15) The Book of Why: The New Science of Cause and Effect, Basic Books, (my note).

3

Koller & Milch (2003) Multi-Agent Influence Diagrams for Representing and Solving Games, Games and economic behavior, https://www.sciencedirect.com/science/article/pii/S0899825602005444.

4

Castelletti, Consonni & Vedova et al. (2018-12) Learning Markov Equivalence Classes of Directed Acyclic Graphs: An Objective Bayes Approach, Bayesian Analysis, https://projecteuclid.org/journals/bayesian-analysis/volume-13/issue-4/Learning-Markov-Equivalence-Classes-of-Directed-Acyclic-Graphs--An/10.1214/18-BA1101.full.

5

Bergemann & Morris (2016-05) Bayes Correlated Equilibrium and the Comparison of Information Structures in Games: Bayes Correlated Equilibrium, Theoretical Economics, http://doi.wiley.com/10.3982/TE1808.

6

Kamenica & Gentzkow (2011-10) Bayesian Persuasion, American Economic Review, https://www.aeaweb.org/articles?id=10.1257/aer.101.6.2590.