default Bayesian factors

Terms from Statistics for HCI: Making Sense of Quantitative Data

Bayesian statistics requires some form of prior. Sometimes this is based on past data or is in the form of a researcher's beliefs encoding (belief) as probability, but where there are no clear grounds for either of these, default Bayesian factors may be used. Typically this will be a uniform distribution over bounded values: for example, assigning a prior probability of 1/6 for each of the potential faces of a die. In the case of unbounded data the Cauchy distribution is a frequent default choice although this also requires a central value and spread parameter.

Used on page 77