Bayesian experimental design
From Wikipedia, the free encyclopedia
Bayesian experimental design differs from the classical approach in that the purpose of the experiment is explicitly represented in the form of a loss function. Different loss functions imply different ways to optimise the design. Designing to best estimate model parameters leads to Bayes a-optimal designs, whereas designing to maximise the information gained leads to Bayes d-optimal designs.
[edit] References
- Chaloner and Verdinelli Bayesian Experimental Design: A Review Statist. Sci., vol 10, 273-304, 1995.
