Decision analysis compares actions by combining uncertain outcomes with their probabilities and assigned utilities or losses. It extends Bayesian inference: a posterior probability describes uncertainty, but does not by itself determine what to do. section 10.2.6
Choose the act with the greatest expected utility using the evidence available at the time. Decision trees organize actions, uncertain events, and consequences; rollback evaluates the choices. Sensitivity analysis tests whether changed assumptions alter the preferred act, and value of information weighs an additional test or experiment against its cost. section 7.5.1 section 10.2.6
Medical treatment
A physician weighing treatment, no treatment, or another test must consider the posterior probability alongside side effects, disease consequences, and the cost and accuracy of further information. section 10.2.6
Staged R&D investment
A company can fund a prototype, update its estimates from the results, then choose whether to continue, redesign, abandon, or run another experiment. Reassessing each commitment limits exposure without treating past spending as a reason to continue. section 7.5.1
