Abstract:
|
In drug trials there may be interest in estimating the hypothetical effect of the drug under conditions of perfect adherence. Those are conditions, of course, which did not obtain in the trial itself. Care is required, but often lacking, in explaining what the hypothetical conditions are. I will discuss several versions of the alternate universe in which adherence was perfect. The effects of the drug may be different in these different universes, so that no single estimator can consistently estimate all of them. Furthermore, estimators commonly claimed to estimate de jure effects do not consistently estimate any of them.
|