Sample size calculations based on slopes and other summary statistics

Abstract

Sample size calculations based on two-sample comparisons of slopes have been reported by many. This paper extends such discussions to include summary statistics other than slopes, such as post-baseline means, change scores, and final observations. Specifically, sample size formulas for analyses based on a broad class of summary statistics are presented, with modifications proposed to allow for missing data caused by staggered entry and random dropouts. The formulas developed are used to illustrate how required sample size is affected by summary statistic choice, variance parameters, the type of treatment difference of interest, and the manner in which incomplete observations are used in the analysis. An example based on longitudinal data from the Muscatine Study is presented.