Strengthening experimental design by balancing potentially confounding variables across treatment groups.
Strengthening experimental design by balancing potentially confounding variables across treatment groups.
复制标题
通过平衡治疗组之间潜在的混淆变量来加强实验设计。
DOI：
10.1097/00006199-199711000-00009
复制
发表时间：
1997
期刊：
影响因子：
2.5
通讯作者：
Zeller,DL
中科院分区：
文献类型：
--
作者：
Zeller,R;Good,M;Anderson,GC;Zeller,DL
将受试者简单随机分配到治疗组和对照组为实验研究提供了两个主要好处：（a）各组之间已知和未知受试者特质的平衡和（B）治疗效果和随机误差变化的无偏估计。然而，简单的随机分配并不能保证各组具有可比性。例如，使用简单随机分配（以下简称随机分配）的研究存在产生在一个或多个水平的混杂协变量中不平衡的组的风险（Silverman，1985）。这在小型研究中尤其如此，但在大型研究中也是一个值得关注的问题（Cook &坎贝尔，1979）。最小化是一种分配给组的方法，通过控制尽可能多的已知无关变异源来减少偏倚量。它产生的群体是可比的规模和分布的潜在混淆协变量（Pocock，1979年; Pocock & Simon，1975年;白色& Freedman，1978年）。这篇文章是为护士研究人员谁是设计随机临床试验，无论是作为他们的研究计划成熟和测试护理干预措施的需求增加。在Conlon和安德森（1990）工作的基础上，本报告包括在干预研究中使用最小化的基本原理，并描述了将受试者分配到多达8个组的最小化-8（min-8）计算机程序的开发和测试。使用最小化的护理研究人员将具有强大的，严格的和可行的研究优势。
Simple random assignment of subjects to treatment and control groups provides two major benefits to experimental research:(a) balance of known and unknown subject idiosyncracies across groups and (b) unbiased estimates of both treatment effects and random error variation. However, simple random assignment does not guarantee that groups will be comparable. For example, studies using simple random assignment (hereafter referred to as random assignment) are at risk for producing groups that are imbalanced in one or more levels of confounding covariates (Silverman, 1985). This is especially true of small studies, but it is also a concern in large ones (Cook & Campbell, 1979). Minimization is a method of assignment to groups that reduces the amount of bias by controlling for as many known extraneous sources of variation as possible. It produces groups that are comparable in size and distribution of potentially confounding covariates (Pocock, 1979; Pocock & Simon, 1975; White & Freedman, 1978).This article is written for nurse researchers who are designing randomized clinical trials both as their research programs mature and as the demand for tested nursing interventions increases. Building on the work of Conlon and Anderson (1990), this report includes a rationale for using minimization in intervention studies and describes the development and testing of the minimization-8 (min-8) computer program for assigning subjects to as many as eight groups. Nurse researchers who use minimization will have a powerful, rigorous, and feasible research advantage.