Header

UZH-Logo

Maintenance Infos

Bayesian networks as a tool for epidemiological systems analysis


Lewis, F I (2012). Bayesian networks as a tool for epidemiological systems analysis. In: 9th International Conference on Mathematical Problems in Engineering, Aerospace and Sciences ICNPAA 2012, Vienna, 10 July 2012 - 14 July 2012, 610-617.

Abstract

Bayesian network analysis is a form of probabilistic modeling which derives from empirical data a directed acyclic graph (DAG) describing the dependency structure between random variables. Bayesian networks are increasingly finding application in areas such as computational and systems biology, and more recently in epidemiological analyses. The key distinction between standard empirical modeling approaches, such as generalised linear modeling, and Bayesian network analyses is that the latter attempts not only to identify statistically associated variables, but to additionally, and empirically, separate these into those directly and indirectly dependent with one or more outcome variables. Such discrimination is vastly more ambitious but has the potential to reveal far more about key features of complex disease systems. Applying Bayesian network modeling to biological and medical data has considerable computational demands, combined with the need to ensure robust model selection given the vast model space of possible DAGs. These challenges require the use of approximation techniques, such as the Laplace approximation, Markov chain Monte Carlo simulation and parametric bootstrapping, along with computational parallelization. A case study in structure discovery - identification of an optimal DAG for given data - is presented which uses additive Bayesian networks to explore veterinary disease data of industrial and medical relevance.

Abstract

Bayesian network analysis is a form of probabilistic modeling which derives from empirical data a directed acyclic graph (DAG) describing the dependency structure between random variables. Bayesian networks are increasingly finding application in areas such as computational and systems biology, and more recently in epidemiological analyses. The key distinction between standard empirical modeling approaches, such as generalised linear modeling, and Bayesian network analyses is that the latter attempts not only to identify statistically associated variables, but to additionally, and empirically, separate these into those directly and indirectly dependent with one or more outcome variables. Such discrimination is vastly more ambitious but has the potential to reveal far more about key features of complex disease systems. Applying Bayesian network modeling to biological and medical data has considerable computational demands, combined with the need to ensure robust model selection given the vast model space of possible DAGs. These challenges require the use of approximation techniques, such as the Laplace approximation, Markov chain Monte Carlo simulation and parametric bootstrapping, along with computational parallelization. A case study in structure discovery - identification of an optimal DAG for given data - is presented which uses additive Bayesian networks to explore veterinary disease data of industrial and medical relevance.

Statistics

Altmetrics

Downloads

31 downloads since deposited on 04 Feb 2013
22 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Conference or Workshop Item (Paper), refereed, original work
Communities & Collections:05 Vetsuisse Faculty > Chair in Veterinary Epidemiology
Dewey Decimal Classification:570 Life sciences; biology
610 Medicine & health
Language:English
Event End Date:14 July 2012
Deposited On:04 Feb 2013 12:47
Last Modified:07 Dec 2017 18:42
Publisher:American Institute of Physics
Series Name:AIP Conference Proceedings
Number:1493
ISSN:0094-243X
Free access at:Publisher DOI. An embargo period may apply.
Publisher DOI:https://doi.org/10.1063/1.4765550
Related URLs:http://icnpaa.com/index.php/icnpaa/2012

Download

Download PDF  'Bayesian networks as a tool for epidemiological systems analysis'.
Preview
Content: Published Version
Filetype: PDF
Size: 334kB
View at publisher