Bayesian fitting procedures for hydrological point processes.

Parry, K

Bayesian fitting procedures for hydrological point processes.

aut.researcher	Parry, Katharina
dc.contributor.author	Parry, K
dc.date.accessioned	2015-01-29T23:47:38Z
dc.date.available	2015-01-29T23:47:38Z
dc.date.copyright	2014-11-27
dc.date.issued	2014-11-27
dc.description.abstract	Model fitting and selection typically requires the use of likelihoods. Applying standard methods to hydrological point processes, however, is problematic as their likelihoods are often analytically intractable and the data sets used for analysis are very large. We consider the use of Approximate Bayesian Computation (ABC) to fit these models without the need to calculate the likelihood, in conjunction with the Deviance Information Criterion (DIC) for model selection. ABC works by simulating artificial data for different parameter values, and comparing the summary statistics of the simulated data to the summary statistics of the observed data. A critical problem is that ABC only works well in lower dimensions, that is, no more than three or four summary statistics should be used. In addition, the choice of the set of statistics is relevant for the accuracy of the parameter estimation. In this presentation, we discuss the process of finding a suitable subset of summary statistics. This work has important applications for the use of ABC in hydrological modelling.
dc.identifier.citation	2014 AUT MATHEMATICAL SCIENCES SYMPOSIUM held at AUT, Auckland, New Zealand, 2014-11-27to 2014-11-28
dc.identifier.uri	https://hdl.handle.net/10292/8357
dc.rights.accessrights	OpenAccess
dc.title	Bayesian fitting procedures for hydrological point processes.
dc.type	Conference Contribution
pubs.elements-id	178605
pubs.organisational-data	/AUT
pubs.organisational-data	/AUT/Design & Creative Technologies