In order to use Bayesian inference, we need to have a prior distribution representing beliefs and uncertainties about unknown quantities before the data are observed. This will often represent the prior beliefs of a client or fellow researcher. Elicitation (O’Hagan et al. 2006) is the process of extracting knowledge, beliefs, and uncertainties about unknown quantities from the client so that these can be expressed as a prior probability distribution. This can sometimes be a difficult task, especially, but not only, when the client has had little training in probability. It is known that people can have a tendency to over- or underestimate probabilities under various circumstances and to make other errors so elicitation questions should be carefully designed.
- O’Hagan A, Buck CE, Daneshkhah A, Eiser JR, Garthwaite PH, Jenkinson DJ, Oakley JE, Rakov T (2006) Uncertain judgements: eliciting experts’ probabilities. Wiley, ChichesterGoogle Scholar