Jeremy Oakley & Paul Garthwaite

Knowledge Elicitation

Venue: Harrington Building (room HA129), University of Central Lancashire, Preston.

Date: Thursday 06-12-2018, 2 - 5pm

Jeremy Oakley (University of Sheffield)

Paul Garthwaite (Open University)

Prof. Jeremy Oakley (School of Mathematics and Statistics, University of Sheffield)

Title: "Eliciting probability distributions with SHELF"

Abstract: Eliciting a probability distribution is the process of extracting an expert's (or group of experts') knowledge about some uncertainty quantity of interest, and choosing a suitable distribution to represent the expert's uncertainty. I will describe a general approach for elicitation: the Sheffield Elicitation Framework (SHELF). SHELF is a behavioural aggregation method for eliciting a distribution from a group of experts. The process involves a face-to-face discussion between the experts, managed by a facilitator. Individual opinions from all participants are first recorded, and then a single distribution is chosen following debate and discussion. Particular emphasis is placed on justifying the chosen distribution in relation to available, documented evidence. I will also demonstrate various (freely available) software tools that can be used to support the process.

Prof. Paul Garthwaite (Department of Mathematics and Statistics, Open University)

Title: "Quantifying expert opinion about some sampling models that include covariates"

Abstract: This talk will describe some elicitation methods for quantifying expert opinion in the contexts of regression, categorical data and generalized linear models. The methods all use interactive graphics to elicit assessments from the expert and are implemented in (free) software. The aim is to ask the expert to perform meaningful tasks and convert the expert’s assessments into a probability distribution that gives a useful representation of his or her opinions. Examples where methods have been used will be given. One example concerns a treatment pathway model that was developed to examine the costs and benefits of the current bowel cancer service in England and to evaluate potential alternatives in service provision. To use the pathway model, various parameters and probability distributions had to be specified. They could not all be determined from empirical evidence and, instead, expert opinion was elicited in the form of probability distributions that gave the required information.

The programme of the seminar is as follows:

2 - 3 pm -- Prof. Jeremy Oakley

3 - 3.20 pm -- Refreshments

3.20 - 4.20 pm -- Prof. Paul Garthwaite

4.20 Examples of knowledge elicitation for UCLan ICONS project and Discussion (Dr. Svetlana Tishkovskaya)