Bootstrap Analysis Double-Independent Programming: Issues and Solutions.
Nils Pénard · 2012
Bootstrap analysis is a valuable, iterative, statistical analysis tool useful when the distribution of an analysis variable is unknown, and likely non-normal and/or non-symmetrical. This family of techniques is garnering attention in the pharmaceutical industry with the development of faster and more powerful computers. The downside of such iterative analysis emerges when the final bootstrap results are to be included in a submission to regulatory authorities. In order to ensure that the bootstrap results are validated appropriately, double-independent programming is the preferred method. Unfortunately, bootstrap ADaM datasets used for these planned analyses are large, complex, and challenging to produce. They also cause extended processing times reading or writing to storage. Doubleprogramming only magnifies these issues. This paper will present a case study of bootstrap analysis with a PRO efficacy endpoint to illustrate the issues and solutions that arise during double-independent programming validation.