Estimating Mediation Effects in Prevention Studies

摘要 这项相互竞争的延续赠款提案的目的是继续开发、评估和 应用方法学和统计程序来确定预防方案如何改变结果 变量。这些中介分析评估程序对 预防计划和对结果的影响。正如许多研究人员指出的那样，调解分析确定了 最有效的计划组件，并增加对基础流程的了解 不断变化的结果变量。 这笔赠款的私人投资者获得了一笔为期一年的NIDA小额赠款和两笔为期四年的赠款，以开发和 对调解分析方法和统计程序的设计和程序进行一般评估 用于预防研究。拟议的五年延续研究侧重于以下方面的实际需要 研究人员，将最新的统计发展应用于中介分析，并开发新的方法 预防研究设计。四个统计主题代表了这项研究的下一步，将包括 分析和模拟研究以及对实际病因学和预防数据的应用。在研究1中， 对显著性检验中的实际问题进行了研究，包括对有限条件下研究的最佳方法。 样本量和评估任何调解设计所需样本量的一般方法。研究2展开 评估违反调解模型假设的方法如何导致不正确的结论。学习 2将因果推理的新发展扩展到中介分析，以调查和检验以下假设 调解分析。研究3开发和评估超越潜在成长的纵向调解方法 在之前的资助中调查过的曲线建模。调解分析的概念和统计替代方案 来自两个浪和三个或更多浪的数据将被彻底调查。新的纵向替代方案 基于自回归、潜差和指数衰减微分方程的中介模型 模型将被开发和比较。研究4评估了我们先前研究中开发的一般模型 调查调解时，也存在主持人效应或项目效果不同的子组 参与者。缓和效应在预防研究中很常见。这项研究还将包括纵向 一般调解和调节模型中的关系，以及调查在 预防科学。研究5将统计方法应用于NIH资助的几个预防数据 提供有关模型有用性的重要反馈的集合。研究6传播新信息 通过我们的网站、与研究人员的交流和 项目。

Abstract The purpose of this competing continuation grant proposal is to continue developing, evaluating and applying methodological and statistical procedures to determine how prevention programs change outcome variables. These mediation analyses assess the link between program effects on the constructs targeted by a prevention program and effects on the outcome. As noted by many researchers, mediation analyses identify the most effective program components and increase understanding of the underlying processes leading to changing outcome variables. The P. I. of this grant received a one-year NIDA small grant and two four-year grants to develop and evaluate mediation analysis methods and statistical procedures for the designs and procedures commonly used in prevention research. The proposed five-year continuation study focuses on practical needs of researchers, applying recent statistical developments to mediation analysis, and developing methods for new prevention research designs. Four statistical topics represent the next steps in this research and will include analytical and simulation research as well as applications to actual etiological and prevention data. In Study 1, practical issues in significance testing are investigated including the best methods for studies with limited sample size and general methods to assess required sample size for any mediation design. Study 2 develops methods to assess how violations of assumptions of the mediation model lead to incorrect conclusions. Study 2 extends new developments in causal inference to mediation analysis to investigate and test assumptions of mediation analysis. Study 3 develops and evaluates longitudinal mediation methods beyond the latent growth curve modeling investigated in prior funding. Conceptual and statistical alternatives for mediation analysis of data from two waves and three or more waves will be thoroughly investigated. New alternative longitudinal mediation models based on autoregressive, latent difference, and exponential decay differential equation models will be developed and compared. Study 4 evaluates a general model developed in our prior research to investigate mediation when there are also moderator effects or program effects that differ by subgroups of participants. Moderator effects are common in prevention research. This study will also incorporate longitudinal relations in the general mediation and moderation model as well as investigate the types of effects common in prevention science. Study 5 applies the statistical methods to data from several NIH funded prevention data sets providing important feedback about the usefulness of the models. Study 6 disseminates new information about mediation analysis through our website, by communication with researchers, and publications from the project.