Planning for Model Discrimination

Early strategies for model discrimination called for experiments at locations where the predictions of the leading candidate models differed most. Such a criterion was given by Hunter and Reiner (1965) for discrimination between 

Box and Hill (1967) gave a more natural criterion that the next event be designed for maximum expected information gain (entropy decrease). They used the entropy function 

Here the indices i and m used by Box and Hill are replaced by j and nmod, to save them for other use in Chapter 7. In their design criterion, Box and Hill actually used an upper bound D on the entropy decrease, which is more easily computed than the expectation R itself. 

The observation was regarded as normally distributed, with a known variance cr bility density 

---