ABSTRACT
Bayesian inference is one of the two dominant statistical philosophies. It differs from frequentism in many important ways, such as permitting the analyst (1) to express uncertainty as a probability, (2) to integrate over the parameter space, and (3) to take account of prior information or opinion. Until 1990, most realistic problems were intractable for Bayesian analysis, but then Markov Chain Monte Carlo (MCMC) methods revolutionized the field. Nearly all statisticians now accept the value of Bayesian methods, even if they are not used for all applications. This chapter is intended as a gentle and mostly non-technical introduction to Bayesian thinking and tools in a forensic context. Two toy applications are used for motivation: an examination of possibly fraudulent insurance claims, and an investigation into the possibility that a hospital nurse is a serial murderer.
