ABSTRACT
In Chapter 2 we have described adaptive allocation methods based on previous treat-
ment allocations. However, it is intuitively very appealing to choose progressive as-
signments on the basis of the observed outcomes as well. In some cases, this might
prove a necessity, for instance
i) in local optimality problems, when the target allocation is a priori unknown due
to its dependence on unknown quantities in the statistical model, such as the pa-
rameters, on the functional form of the expected response, etc.;
ii) with ethical concerns about the risks for the subjects involved in the trial, or to
economic concern for costs.