How Should One Choose Stage Sizes?

The literature has few guidelines, even for 2-stage procedures for product of means problem.

Question

For a 2-stage rule with N=100, what should be?

Notice that Rekab's asymptotic results are of little help to a practitioner faced with this question. Further, his results are typical of the results that one finds in the literature concerning few-stage designs.

Noble is somewhat unusual, in that he gives more concrete guidelines. From Noble, if

Our results for uniform priors show that the optimal answer is

Rough calculations indicate that one would need at least N=10,000 before Noble's bounds would apply to uniform priors.

Stage Size Growth Outline Efficiency