Nguyen Phuc H, Herring Amy H, Engel Stephanie M
Department of Statistical Science, Duke University, Durham, 27710, NC, USA.
Department of Epidemiology, The University of North Carolina at Chapel Hill, Chapel Hill, 27516, NC, USA.
Stat Biosci. 2024 Jul;16(2):321-346. doi: 10.1007/s12561-023-09385-7. Epub 2023 Oct 1.
Estimating sample size and statistical power is an essential part of a good epidemiological study design. Closed-form formulas exist for simple hypothesis tests but not for advanced statistical methods designed for exposure mixture studies. Estimating power with Monte Carlo simulations is flexible and applicable to these methods. However, it is not straightforward to code a simulation for non-experienced programmers and is often hard for a researcher to manually specify multivariate associations among exposure mixtures to set up a simulation. To simplify this process, we present the R package mpower for power analysis of observational studies of environmental exposure mixtures involving recently-developed mixtures analysis methods. The components within mpower are also versatile enough to accommodate any mixtures methods that will developed in the future. The package allows users to simulate realistic exposure data and mixed-typed covariates based on public data set such as the National Health and Nutrition Examination Survey or other existing data set from prior studies. Users can generate power curves to assess the trade-offs between sample size, effect size, and power of a design. This paper presents tutorials and examples of power analysis using mpower.
估计样本量和统计功效是良好的流行病学研究设计的重要组成部分。简单假设检验有封闭式公式，但针对暴露混合物研究设计的先进统计方法却没有。用蒙特卡罗模拟估计功效很灵活，适用于这些方法。然而，对于没有经验的程序员来说，编写模拟代码并非易事，而且研究人员通常很难手动指定暴露混合物之间的多变量关联来建立模拟。为简化这一过程，我们推出了R包mpower，用于对涉及最近开发的混合物分析方法的环境暴露混合物观察性研究进行功效分析。mpower中的组件也足够通用，能够适应未来将开发的任何混合物方法。该软件包允许用户基于公共数据集（如国家健康与营养检查调查）或先前研究的其他现有数据集，模拟现实的暴露数据和混合型协变量。用户可以生成功效曲线，以评估样本量、效应量和设计功效之间的权衡。本文介绍了使用mpower进行功效分析的教程和示例。
