|
[Print-friendly version]
 Rebecca O’Leary
Faculty of Science, Statistical Science and Operations Research
Thesis Title: Expert Elicitation in Ecology for Bayesian Modelling: The Role of the Scientist in the Decision Making Process
Current Thesis Abstract: I am doing a PhD in expert elicitation in ecology, in particular predicting the habitat suitability of rare or endangered Australian species. Modelling the habitat suitability of these species has become increasingly important for conservation and wildlife management. However, modelling these species can be problematic, since the dataset usually contains limited presence sites (Engler et al. 2004; Manel et al. 1999) and may contain mostly of presences or absences (Pearce et al. 2001; Radeloff et al. 1998). Expert opinion has been acknowledged as having some value in complementing the small datasets on which such models are typically built.
We developed a simple elicitation approach for single or multiple experts. Each expert's opinion is a rating of increasing, decreasing or zero value indicating species response to a covariate, and also elicit the experts' confidence. These opinions were formulated into a mixture of three normally distributed priors and combined with the limited presence/absence data, modelled via logistic regression. This approach is illustrated using the threatened Australian brush-tailed rock-wallaby
(Petrogale penicillata) and two experts.
Another project is a comparison of three expert elicitation methods for the logistic regression. These methods are the graphically-assisted predictive approach (Kynn 2006); geographically-assisted predictive method (Denham & Mengersen 2006); and
simple a direct questionnaire elicitation approach (mentioned above). These methods were trialled on two experts in order to model the species distribution of the threatened brush tailed rock wallaby.
|
