Size of Interventional Markov Equivalence Classes in Random DAG Models

Directed acyclic graph (DAG) models are popular for capturing causal relationships. From observational and interventional data, a DAG model can only be determined up to its \emph{interventional Markov equivalence class} (I-MEC). We investigate the size of MECs for random DAG models generated by uniformly sampling and ordering an Erdős-Renyi graph. For constant density, we show that the expected $\log$ observational MEC size asymptotically (in the number of vertices) approaches a constant. We characterize I-MEC size in a similar fashion in the above settings with high precision. We show that the asymptotic expected number of interventions required to fully identify a DAG is a constant. These results are obtained by exploiting Meek rules and coupling arguments to provide sharp upper and lower bounds on the asymptotic quantities, which are then calculated numerically up to high precision. Our results have important consequences for experimental design of interventions and the development of algorithms for causal inference.

[1]  Michael D. Perlman,et al.  Enumerating Markov Equivalence Classes of Acyclic Digraph Models , 2001, UAI.

[2]  M. Gangl.  Causal Inference in Sociological Research , 2010 .

[3]  S. Gillispie.  Formulas for counting acyclic digraph Markov equivalence classes , 2006 .

[4]  Xia Li,et al.  Gene Perturbation Atlas (GPA): a single-gene perturbation repository for characterizing functional mechanisms of coding and non-coding genes , 2015, Scientific Reports.

[5]  Jean Honorio,et al.  Computationally and statistically efficient learning of causal Bayes nets using path queries , 2017, NeurIPS.

[6]  Bin Yu,et al.  Formulas for Counting the Sizes of Markov Equivalence Classes of Directed Acyclic Graphs , 2016, ArXiv.

[7]  Christopher Meek,et al.  Causal inference and causal explanation with background knowledge , 1995, UAI.

[8]  David Sontag,et al.  SparsityBoost: A New Scoring Function for Learning Bayesian Network Structure , 2013, UAI.

[9]  Caroline Uhler,et al.  Learning directed acyclic graphs based on sparsest permutations , 2013, ArXiv.

[10]  Jean Honorio,et al.  Learning causal Bayes networks using interventional path queries in polynomial time and sample complexity , 2017, 1706.00754.

[11]  Massimiliano Pontil,et al.  Empirical Bernstein Bounds and Sample-Variance Penalization , 2009, COLT.

[12]  Yuhao Wang,et al.  Permutation-based Causal Inference Algorithms with Interventions , 2017, NIPS.

[13]  Robert Castelo,et al.  On Inclusion-Driven Learning of Bayesian Networks , 2003, J. Mach. Learn. Res..

[14]  Caroline Uhler,et al.  Counting Markov Equivalence Classes by Number of Immoralities , 2016, UAI.

[15]  Bertran Steinsky,et al.  Enumeration of labelled chain graphs and labelled essential directed acyclic graphs , 2003, Discret. Math..

[16]  Peter Bühlmann,et al.  Two Optimal Strategies for Active Learning of Causal Models from Interventions , 2012, ArXiv.

[17]  Ioannis Tsamardinos,et al.  Probabilistic Computational Causal Discovery for Systems Biology , 2016 .

[18]  Peter Bühlmann,et al.  Characterization and Greedy Learning of Interventional Markov Equivalence Classes of Directed Acyclic Graphs (Abstract) , 2011, UAI.

[19]  Caroline Uhler,et al.  Counting Markov equivalence classes for DAG models on trees , 2017, Discret. Appl. Math..

[20]  Stephan G. Wagner.  Asymptotic Enumeration of Extensional Acyclic Digraphs , 2012, Algorithmica.

[21]  Adrian Vetta,et al.  Randomized Experimental Design for Causal Graph Discovery , 2014, NIPS.

[22]  Daphne Koller,et al.  Ordering-Based Search: A Simple and Effective Algorithm for Learning Bayesian Networks , 2005, UAI.

[23]  Frederick Eberhardt,et al.  Experiment selection for causal discovery , 2013, J. Mach. Learn. Res..

[24]  P. Spirtes,et al.  Causation, prediction, and search , 1993 .

[25]  Caroline Uhler,et al.  Consistency Guarantees for Permutation-Based Causal Inference Algorithms , 2017 .

[26]  Alexandros G. Dimakis,et al.  Cost-Optimal Learning of Causal Graphs , 2017, ICML.

[27]  Frederick Eberhardt,et al.  On the Number of Experiments Sufficient and in the Worst Case Necessary to Identify All Causal Relations Among N Variables , 2005, UAI.

[28]  Peter Bühlmann,et al.  Estimating High-Dimensional Directed Acyclic Graphs with the PC-Algorithm , 2007, J. Mach. Learn. Res..

[29]  Caroline Uhler,et al.  Characterizing and Learning Equivalence Classes of Causal DAGs under Interventions , 2018, ICML.

[30]  Charles Wang,et al.  Generalized Permutohedra from Probabilistic Graphical Models , 2016, SIAM J. Discret. Math..

[31]  C. Meek,et al.  Graphical models: selecting causal and statistical models , 1997 .

[32]  Alexandros G. Dimakis,et al.  Learning Causal Graphs with Small Interventions , 2015, NIPS.