A0196
Title: Order-based structure learning without score equivalence
Authors: Hyunwoong Chang - Texas A&M University (United States) [presenting]
Abstract: Structure learning of directed acyclic graph (DAG) models is the task of discovering the underlying DAG structure that represents the conditional independence relations among variables in a given observational data. An empirical Bayes formulation of the structure learning is proposed, where the prior specification assumes that all node variables have the same error variance, an assumption known to ensure the identifiability of the underlying causal DAG. To facilitate efficient posterior computation, the posterior probability of each ordering is approximated by that of a best DAG model, which naturally leads to an order-based Markov chain Monte Carlo (MCMC) algorithm. Strong selection consistency for the model in high-dimensional settings is proved under a condition that allows heterogeneous error variances, and the mixing behavior of the sampler is theoretically investigated. The method is demonstrated to outperform other state-of-the-art algorithms under various simulation settings, and a single-cell real-data study is provided to illustrate the practical advantages of the proposed method.
