Internal Pilot Designs with Compound Symmetric and Gaussian Repeated Measures

• Activity Number: 220
• Type: Contributed
• Date/Time: Tuesday, August 10, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Christopher S. Coffey*+ and Keith E. Muller and Matthew J. Gurka
• Companies: University of Alabama, Birmingham and University of North Carolina, Chapel Hill and University of North Carolina, Chapel Hill
• Keywords: exchangeable correlation ; cluster sample ; univariate approach

Abstract

An internal pilot design uses an interim sample size analysis, without interim data analysis, to adjust the final total number of observations. The approach typically gives a sample size sufficiently large (to achieve the statistical power desired), but not too large (which would waste money and time). Research to assess age differences in cerebral vascular tortuosity (curvature in three dimensions) would benefit greatly from internal pilot designs. However, interest in repeated measures across brain regions precludes using existing univariate methods for internal pilots. We report new methods to meet the need for Gaussian data with constant variance and correlation (compound symmetry) across repeated measures, and no missing observations, as in the tortuosity research. We directly extend all known exact (small sample) univariate methods to repeated measures with compound symmetry and any between subject design with fixed predictors (not just t tests). The new methods allow avoiding sample sizes that are too small or too large, while still controlling false positives with compound symmetric data.