Dawson J D
Department of Preventive Medicine and Environmental Health, University of Iowa, Iowa City 52242, USA.
Biometrics. 1998 Mar;54(1):323-30.
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.
许多人已经报道了基于斜率的两样本比较的样本量计算方法。本文将此类讨论扩展到包括斜率以外的汇总统计量，如基线后均值、变化分数和最终观测值。具体而言，给出了基于一大类汇总统计量进行分析的样本量公式，并提出了修正方法以考虑由交错入组和随机失访导致的缺失数据。所推导的公式用于说明所需样本量如何受到汇总统计量选择、方差参数、感兴趣的治疗差异类型以及分析中使用不完全观测值的方式的影响。文中给出了一个基于马斯卡廷研究纵向数据的示例。
