Systematizing the evaluation of partially controlled studies using principal stratifictaion: From theory to practice

We often wish to evaluate treatments and other factors in studies where only some of those factors are directly controlled by the investigators. The framework of "principal stratification" has been proposed recently for evaluating partially controlled studies under such broader settings (Frangakis and Rubin (2002)). We have developed a software package, "PSpack", with appropriate documentation containing examples for implementing principal stratification. We hope that this helps bridge theory with practice, and that it systematizes the use of principal stratification in evaluating partially controlled studies. In this article we provide an introduction to using the software we have developed, which is available with documentation at the web site: http://biosun01.biostat.jhsph.edu/~cfrangak/papers/ps.html.