Abstract

Experimental tests of interventions need to have sufficient sample size to constitute a robust test of the intervention’s effectiveness with reasonable precision and power. To estimate the required sample size adequately, researchers are required to specify an effect size. But what effect size should be used to plan the required sample size? Various inroads into selecting the a priori effect size have been suggested in the literature—including using conventions, prior research, and theoretical or practical importance. In this paper, we first discuss problems with some of the proposed methods of selecting the effect size for study planning. We then lay out a method for intervention researchers that provides a way out of many of these problems. The proposed method requires setting a meaningful change definition, it is specifically suited for applied researchers interested in planning tests of intervention effectiveness. We provide a hands-on walk through of the method and provide easy-to-use R functions to implement it.

摘要

干预的实验测试需要有足够的样本量，以构成对干预有效性的稳健测试，并具有合理的精度和效力。为了充分估计所需的样本量，研究人员需要指定效应量。但是应该使用什么效应量来规划所需的样本量？选择先验的各种方法文献中已经提出了效应量——包括使用惯例、先前的研究以及理论或实践的重要性。在本文中，我们首先讨论了一些为研究计划选择效应大小的建议方法的问题。然后，我们为干预研究人员制定了一种方法，该方法提供了解决许多这些问题的方法。所提出的方法需要设置一个有意义的变化定义，它特别适合对计划干预测试感兴趣的应用研究人员。我们提供了该方法的实践演练，并提供了易于使用的R函数来实现它。