A0979
Title: Efficient learning of DAG structures in heavy-tailed data
Authors: Wei Zhou - Southwestern University of Finance and Economics (China) [presenting]
Abstract: Directed acyclic graph (DAG) models are widely used to discover causal relationships among random variables. However, most existing DAG learning algorithms are not directly applicable to heavy-tailed data, which are commonly observed in finance and other fields. The aim is to propose a two-step efficient algorithm based on topological layers, referred to as TopHeat, to learn linear DAGs with heavy-tailed error distributions, which include Pareto, Frechet, log-normal, Cauchy distributions, and so on. First, the topological layers are reconstructed hierarchically in a top-down fashion based on the new reconstruction criteria for heavy-tailed DAGs without assuming the popularly employed faithfulness condition. Second, the directed edges are recovered via the modified conditional independence testing for heavy-tailed distributions. The consistency of the exact DAG structures is theoretically demonstrated. Monte Carlo simulations validate the outstanding finite-sample performance of the proposed algorithm compared with competing methods. In the real data analysis, the exchange rates are analyzed among 17 countries and uncover the source of financial contagion and the pathways, which indicates that the financial risk contagion effect became increasingly stable among European countries as the Euro was introduced.
