Learning Bayesian Networks from Correlated Data

• Activity Number: 313
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Harold Bae* and Stefano Monti and Monty Montano and Thomas T. Perls and Paola Sebastiani
• Companies: Oregon State University and Boston University School of Medicine and Harvard Medical School and Boston University School of Medicine and Boston University
• Keywords: Bayesian networks ; family-based data ; model selection ; parameterization ; mixed models

Abstract

Bayesian networks are probabilistic models that can represent complex distributions in a modular way and have become very popular in many fields. There are many methods to build Bayesian networks from data but a crucial and restrictive assumption underlying all proposed approaches is that data are a random sample of independent and identically distributed observations. However, many observational studies are designed using some form of clustered sampling that introduces correlations between observations within the same cluster, and there are no methods to learn Bayesian networks from correlated data; the inflated Type I errors due to ignoring the correlation in the data will result in highly connected networks. We describe a novel parameterization of Bayesian networks that uses random effects to model the correlation within sample units while maintaining the decomposability of the likelihood and can be used for structure and parameter learning of Bayesian networks from correlated data. We compare different learning metrics using simulations and illustrate the method in a real example of genetic and non-genetic factors associated with human longevity.