How big does my sample need to be? A primer on the murky world of sample size estimation
How big does my sample need to be? A primer on the murky world of sample size estimation
复制标题
DOI：
10.1016/j.ptsp.2005.05.004
复制
发表时间：
2005-08-01
影响因子：
2.4
通讯作者：
Atkinson, G
中科院分区：
文献类型：
--
作者：
Batterham, AM;Atkinson, G
背景：现在，大多数提交给资助机构、伦理委员会的提案，以及越来越多的提交给期刊发表的文章，都必须明确说明样本量的理由。然而，样本大小估计的过程往往令人困惑。目的：在这里，我们提出了一个样本量估计的入门，试图揭开这个过程的神秘面纱。方法：首先，我们讨论了功率分析和样本量估计中涉及的参数。这些包括最小的值得检测的影响，类型I和II错误率，和可变性的结果测量。其次，通过一个简化的示例“对话”，我们说明了分配适当参数值以达到所需样本量的大致数字所涉及的决策过程。我们采用假设、平行组、随机试验设计，但一般原则和概念可转移到其他设计。该说明是基于传统的，功率分析，零假设检验框架。简而言之，我们还讨论了基于平均效应估计所需精度的样本量估计方法。结论：严格的样本量规划是重要的。研究人员应该诚实和明确地考虑到在样本量估计中涉及的每个参数所做的决定。(c) 2005 Elsevier Ltd版权所有。
Background: An explicit justification of sample size is now mandatory for most proposals submitted to funding bodies, ethics committees and, increasingly, for articles submitted for publication in journals. However, the process of sample size estimation is often confusing.Aim: Here, we present a primer of sample size estimation in an attempt to demystify the process.Method: First, we present a discussion of the parameters involved in power analysis and sample size estimation. These include the smallest worthwhile effect to be detected, the Types I and II error rates, and the variability in the outcome measure. Secondly, through a simplified, example 'dialogue', we illustrate the decision-making process involved in assigning appropriate parameter values to arrive at a ballpark figure for required sample size. We adopt a hypothetical, parallel-group, randomized trial design, though the general principles and concepts are transferable to other designs. The illustration is based on a traditional, power-analytic, null hypothesis-testing framework. In brief, we also address sample size estimation methods based on the required precision of the mean effect estimate. Conclusion: Rigorous sample size planning is important. Researchers should be honest and explicit regarding the decisions made for each of the parameters involved in sample size estimation. (c) 2005 Elsevier Ltd. All rights reserved.