Shaffer J W
J Behav Med. 1979 Sep;2(3):221-38. doi: 10.1007/BF00844920.
Studies comparing different treatments over time using repeated measures on a wide variety of outcome variables are very common in the research literature. Unfortunately, they are rarely analyzed appropriately, at least from the standpoint of controlling the type 2 error rate (false rejection of the null hypothesis). As a consequence, many of the positive findings reported in such studies may be incapable of replication. The present article reviews the logical and statistical problems involved, emphasizes the importance of controlling the experimentwise error rate regardless of the research design employed, and presents an integrated, multivariate strategy for dealing with the several issues raised. Resolution of these issues is essential if the "fictional" content of research journals is to be kept within reasonable bounds.
随着时间推移，使用重复测量法对各种结果变量进行不同治疗比较的研究在研究文献中非常普遍。不幸的是，这些研究很少得到恰当分析，至少从控制Ⅱ类错误率（错误地拒绝原假设）的角度来看是这样。因此，此类研究中报告的许多阳性结果可能无法重复。本文回顾了其中涉及的逻辑和统计问题，强调了无论采用何种研究设计，控制实验性错误率的重要性，并提出了一种综合的多变量策略来处理所提出的几个问题。如果要将研究期刊中的“虚构”内容控制在合理范围内，解决这些问题至关重要。
