Statistical problems in pharmaceutics are global, recurrent, and
scientifically relatively well defined. Examples of three such problems
Bioequivalence: Is a generic version practically equivalent to Prozac?
Dose-response: What is the therapeutic window for Claritin?
Multiple endpoints: Does Roloxifene reduce breast cancer rate and MI (heart attack) rate for post-menopausal women?
Regulatory decisions granting approval are supposed to be made in such a way that at most 1 decision out of 20 is wrong, in the long run. I will describe a ``new'' statistical principle (called partitioning) for making sound decisions in problems that involve ``multiplicity'' such as these. As I describe these problems scientifically, you are invited to formulate them statistically for/with me (bait), to check the appropriateness of what we teach or are taught, and to participate in the on-going formulation of the multiple endpoints problem (hook).