Oklahoma State University.
Calvin College.
Child Dev. 2018 Nov;89(6):2038-2050. doi: 10.1111/cdev.13097. Epub 2018 May 24.
To evaluate and improve the validity of causal inferences from meta-analyses of longitudinal studies, two adjustments for Time-1 outcome scores and a temporally backwards test are demonstrated. Causal inferences would be supported by robust results across both adjustment methods, distinct from results run backwards. A systematic strategy for evaluating potential confounds is also introduced. The methods are illustrated by assessing the impact of spanking on subsequent externalizing problems (child age: 18 months to 11 years). Significant results indicated a small risk or a small benefit of spanking, depending on the adjustment method. These meta-analytic methods are applicable for research on alternatives to spanking and other developmental science topics. The underlying principles can also improve causal inferences in individual studies.
为了评估和提高从纵向研究荟萃分析中得出因果推论的有效性，本文展示了两种针对时间 1 结局评分的调整方法和一个回溯性检验。如果两种调整方法的结果都稳健一致，那么因果推论就会得到支持，与回溯性检验的结果明显不同。本文还介绍了一种评估潜在混杂因素的系统策略。这些方法通过评估打屁股对后续外化问题（儿童年龄：18 个月至 11 岁）的影响来举例说明。根据调整方法的不同，显著的结果表明打屁股存在较小的风险或益处。这些荟萃分析方法适用于打屁股替代品和其他发展科学主题的研究。这些基本原则也可以提高个体研究中的因果推论。
