Learning and Sampling of Atomic Interventions from Observations

We study the problem of efficiently estimating the effect of an intervention on a single variable (atomic interventions) using observational samples in a causal Bayesian network. Our goal is to give algorithms that are efficient in both time and sample complexity in a non-parametric setting. Tian and Pearl (AAAI `02) have exactly characterized the class of causal graphs for which causal effects of atomic interventions can be identified from observational data. We make their result quantitative. Suppose P is a causal model on a set $\vec{V}$ of n observable variables with respect to a given causal graph G with observable distribution $P$. Let $P_x$ denote the interventional distribution over the observables with respect to an intervention of a designated variable X with x. Assuming that $G$ has bounded in-degree, bounded c-components ($k$), and that the observational distribution is identifiable and satisfies certain strong positivity condition, we give an algorithm that takes $m=\tilde{O}(n\epsilon^{-2})$ samples from $P$ and $O(mn)$ time, and outputs with high probability a description of a distribution $\hat{P}$ such that $d_{\mathrm{TV}}(P_x, \hat{P}) \leq \epsilon$, and: 1. [Evaluation] the description can return in $O(n)$ time the probability $\hat{P}(\vec{v})$ for any assignment $\vec{v}$ to $\vec{V}$ 2. [Generation] the description can return an iid sample from $\hat{P}$ in $O(n)$ time. We also show lower bounds for the sample complexity showing that our sample complexity has an optimal dependence on the parameters $n$ and $\epsilon$, as well as if $k=1$ on the strong positivity parameter.

[1]  Jim Q. Smith,et al.  Causal analysis with Chain Event Graphs , 2010, Artif. Intell..

[2]  Judea Pearl,et al.  Testing Identifiability of Causal Effects , 1995, UAI.

[3]  Jin Tian,et al.  A general identification condition for causal effects , 2002, AAAI/IAAI.

[4]  Constantinos Daskalakis,et al.  Learning and Testing Causal Models with Interventions , 2018, NeurIPS.

[5]  J. Pearl,et al.  Studies in causal reasoning and learning , 2002 .

[6]  James M. Robins,et al.  Probabilistic evaluation of sequential plans from causal models with hidden variables , 1995, UAI.

[7]  Judea Pearl,et al.  Causal networks: semantics and expressiveness , 2013, UAI.

[8]  Emilija Perkovi'c,et al.  Identifying causal effects in maximally oriented partially directed acyclic graphs , 2019, UAI.

[9]  Judea Pearl,et al.  Identification of Joint Interventional Distributions in Recursive Semi-Markovian Causal Models , 2006, AAAI.

[10]  James J. Heckman,et al.  Econometric Evaluation of Social Programs, Part I: Causal Models, Structural Models and Econometric Policy Evaluation , 2007 .

[11]  D. Rubin,et al.  Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction , 2016 .

[12]  Chandler Squires,et al.  ABCD-Strategy: Budgeted Experimental Design for Targeted Causal Structure Discovery , 2019, AISTATS.

[13]  Alon Orlitsky,et al.  On Learning Distributions from their Samples , 2015, COLT.

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

[15]  J. Pearl Causal diagrams for empirical research , 1995 .

[16]  Duncan J. Watts,et al.  Estimating the Causal Impact of Recommendation Systems from Observational Data , 2015, EC.

[17]  Santtu Tikka,et al.  Enhancing Identification of Causal Effects by Pruning , 2018, J. Mach. Learn. Res..

[18]  Ricard Gavaldà,et al.  Identifiability and transportability in dynamic causal networks , 2016, International Journal of Data Science and Analytics.

[19]  Daniel M. Kane,et al.  Testing Bayesian Networks , 2016, IEEE Transactions on Information Theory.

[20]  Sanjoy Dasgupta,et al.  The Sample Complexity of Learning Fixed-Structure Bayesian Networks , 1997, Machine Learning.

[21]  Arthur Lewbel,et al.  The Identification Zoo: Meanings of Identification in Econometrics , 2019 .

[22]  Tom Burr,et al.  Causation, Prediction, and Search , 2003, Technometrics.

[23]  Ilias Diakonikolas,et al.  Learning Structured Distributions , 2016, Handbook of Big Data.

[24]  Alain Hauser,et al.  Jointly interventional and observational data: estimation of interventional Markov equivalence classes of directed acyclic graphs , 2013, 1303.3216.

[25]  Michael E. Sobel,et al.  Causal Inference in the Social Sciences , 2000 .

[26]  J. Robins A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect , 1986 .

[27]  Manabu Kuroki,et al.  IDENTIFIABILITY CRITERIA FOR CAUSAL EFFECTS OF JOINT INTERVENTIONS , 1999 .

[28]  Peter Spirtes,et al.  Introduction to Causal Inference , 2010, J. Mach. Learn. Res..

[29]  Elias Bareinboim,et al.  Estimating Causal Effects Using Weighting-Based Estimators , 2020, AAAI.

[30]  Raphael Rubin,et al.  Rubin's Pathology: Clinicopathologic Foundations of Medicine. , 2011 .

[31]  Jiji Zhang,et al.  Identification of Conditional Causal Effects under Markov Equivalence , 2019, NeurIPS.

[32]  Andrew A. Renshaw,et al.  Rubin??s Pathology. Clinicopathologic Foundations of Medicine , 2008 .

[33]  Piyush Srivastava,et al.  Stability of Causal Inference , 2016, Conference on Uncertainty in Artificial Intelligence.

[34]  Ronitt Rubinfeld,et al.  On the learnability of discrete distributions , 1994, STOC '94.

[35]  Joseph Y. Halpern Axiomatizing Causal Reasoning , 1998, UAI.

[36]  Jiji Zhang,et al.  Causal Identification under Markov Equivalence: Completeness Results , 2019, ICML.

[37]  Frederick Eberhardt,et al.  Do-calculus when the True Graph Is Unknown , 2015, UAI.

[38]  Karthikeyan Shanmugam,et al.  Characterization and Learning of Causal Graphs with Latent Variables from Soft Interventions , 2019, NeurIPS.

[39]  Jin Tian,et al.  On the Testable Implications of Causal Models with Hidden Variables , 2002, UAI.

[40]  Santtu Tikka,et al.  Simplifying Probabilistic Expressions in Causal Inference , 2018, J. Mach. Learn. Res..

[41]  Arnab Bhattacharyya,et al.  Efficient Distance Approximation for Structured High-Dimensional Distributions via Learning , 2020, NeurIPS.

[42]  Peter A. Thwaites,et al.  Causal identifiability via Chain Event Graphs , 2013, Artif. Intell..

[43]  Santtu Tikka,et al.  Identifying Causal Effects with the R Package causaleffect , 2017, 1806.07161.

[44]  Rosa L. Matzkin Nonparametric identification and estimation of polychotomous choice models , 1993 .