Tian Lili
Department of Biostatistics, University at Buffalo, Buffalo, NY 14214-3000, USA.
Stat Med. 2007 Feb 28;26(5):945-53. doi: 10.1002/sim.2589.
The standardized mean difference has been widely used as the most common index of effect magnitude in many applied fields. In this paper, we propose a novel approach using the concept of generalized variable for the confidence interval estimation and hypothesis testing of standardized mean difference. Furthermore, we extend this approach to compare standardized mean differences between two studies or between two strata. Simulation results demonstrate that the proposed approach can provide confidence intervals with excellent coverage properties and can perform hypothesis testing with satisfactory type-I error control.
标准化均数差已在许多应用领域中广泛用作效应量的最常用指标。在本文中，我们提出了一种使用广义变量概念的新方法，用于标准化均数差的置信区间估计和假设检验。此外，我们将此方法扩展到比较两项研究之间或两个层次之间的标准化均数差。模拟结果表明，所提出的方法可以提供具有出色覆盖特性的置信区间，并且可以在令人满意的I型错误控制下进行假设检验。
