Abstract
We recently proposed two methods for estimating Bayesian networks from high-dimensional non-independent and identically distributed data containing exogenous variables and random effects (Kasza et al., 2012 Kasza, J.E., Glonek, G., Solomon, P. (2012). Estimating Bayesian networks for high-dimensional data with complex mean structure. Aust. New Zealand J. Stat. 54(2): 169–187.
[Crossref], [Web of Science ®], , [Google Scholar]
). The first method is fully Bayesian, and the second is “residual”-based, accounting for the effects of the exogenous variables by utilizing the notion of restricted maximum likelihood. We describe the methods and compare their performance using the Kullback–Leibler divergence, which provides a natural framework for comparing posterior distributions. In applications where the exogenous variables are not of primary interest, we show that the potential loss of information about parameters of interest is typically small.
[Crossref], [Web of Science ®], , [Google Scholar]
). The first method is fully Bayesian, and the second is “residual”-based, accounting for the effects of the exogenous variables by utilizing the notion of restricted maximum likelihood. We describe the methods and compare their performance using the Kullback–Leibler divergence, which provides a natural framework for comparing posterior distributions. In applications where the exogenous variables are not of primary interest, we show that the potential loss of information about parameters of interest is typically small.
Original language | English |
---|---|
Pages (from-to) | 135-152 |
Number of pages | 18 |
Journal | Communications in Statistics: Theory and Methods |
Volume | 44 |
Issue number | 1 |
DOIs | |
Publication status | Published - 2015 |
Externally published | Yes |
Keywords
- Bayesian network
- Exogenous variables
- High-dimensional data
- KullbackLeibler divergence
- Gene regulatory networks
- Variance components