<p>In this article, we describe the analytic design of the coordinated set of outcome-wide analyses used for examining longitudinal associations using data from the Global Flourishing Study (GFS), involving a multinational and multidisciplinary group of scholars. We discuss the benefits of outcome-wide analyses and provide details on controlling for high-dimensional confounders using principal components, accounting for complex sampling designs, imputing missing data, conducting sensitivity analyses for unmeasured confounding, meta-analyzing estimates of associations from across countries, and reporting results. We provide a brief illustrative example of the outcome-wide approach by estimating the association of Wave 1 sense of mastery with a wide range of Wave 2 outcomes. The example illustrates how results can be sensitive to analytic decisions, such as different coding strategies for the predictor and the number of principal components included. We conclude by outlining the major strengths and limitations of the employed methodology.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Coordinated outcome-wide analytic methodology for multi-wave analyses of the global flourishing study

  • R. Noah Padgett,
  • Chris Felton,
  • Matt Bradshaw,
  • Ying Chen,
  • Richard G. Cowden,
  • Eric S. Kim,
  • Renae Wilkinson,
  • Byron R. Johnson,
  • Tyler J. VanderWeele

摘要

In this article, we describe the analytic design of the coordinated set of outcome-wide analyses used for examining longitudinal associations using data from the Global Flourishing Study (GFS), involving a multinational and multidisciplinary group of scholars. We discuss the benefits of outcome-wide analyses and provide details on controlling for high-dimensional confounders using principal components, accounting for complex sampling designs, imputing missing data, conducting sensitivity analyses for unmeasured confounding, meta-analyzing estimates of associations from across countries, and reporting results. We provide a brief illustrative example of the outcome-wide approach by estimating the association of Wave 1 sense of mastery with a wide range of Wave 2 outcomes. The example illustrates how results can be sensitive to analytic decisions, such as different coding strategies for the predictor and the number of principal components included. We conclude by outlining the major strengths and limitations of the employed methodology.