Expected Value of Information
- The use of Bayesian and frequentist probabilistic approaches to uncertainty. Thus, decision-analytic methods like cost-effectiveness analysis usually build on data that are at best only partially relevant, not completely accurate, whose estimated values have large variances, etc. One therefore needs to adjust empirical distributions, use explicit judgmental distributions, or collect new data. In determining whether or not to collect additional data, the expected value of information (EVI) approach, as its name implies, invites the analyst to consider the expected value (in the form of a reduction in opportunity loss) of additional information and the costs of getting it. EVI comes in two forms: global, in which the reduction in opportunity loss from making a decision is estimated across all uncertain parameters ; and partial, in which the reduction in loss relates to getting additional information about a specific parameter. Originated by Raiffa and Schlaiffer (1967).