An initial screening experiment may lead to ambiguous conclusions regard-
ing the factors which are active in explaining the variation of an outcome variable:
thus, adding follow-up runs becomes necessary. To better account for model uncer-
tainty, we propose an objective Bayesian approach to follow-up designs, using prior
distributions suitably tailored to model selection. To select the best follow-up runs,
we adopt a model discrimination criterion based on a weighted average of Kullback–
Leibler divergences between predictive distributions for all possible pairs of models.
Our procedure should appeal to practitioners because it does not require prior specifi-
cations, being fully automatic. When applied to real data, it produces follow-up runs
which better discriminate among factors relative to current methodology.

EB00000004448K

Available

• Objective Bayesian model discrimination in follow-up experimental designs