The paper under discussion (described as “the paper” in this discussion) introduces models for the effect of small studies in multivariate meta analysis. This is done under the working hypothesis that this can be analyzed by studying the effect of the study-specific standard error on the outcomes of interest. That is fine with me. However, I am concerned about the way the data are modeled in the paper and the lack of transparency of the way the data are analyzed. In this discussion, I propose a modification of the model and a simplification of the analysis. I present a more detailed analysis of the two case studies. The P-values obtained in that analysis are close to the ones obtained in the paper. Finally, a sharper correction for multiple testing is presented and the lack of interpretation of “just P-values” is commented.

中文翻译：

---

关于“在多变量荟萃分析中测试小型研究效果”的讨论，作者：Chuan Hong、Georgia Salanti、Sally Morton、Richard Riley、Haitao Chu、Stephen E. Kimmel 和 Yong Chen

正在讨论的论文（在本次讨论中被称为“论文”）介绍了多元荟萃分析中小型研究效果的模型。这是在工作假设下完成的，即可以通过研究特定于研究的标准误差对感兴趣结果的影响来分析这一点。这对我来说没问题。然而，我担心论文中数据建模的方式以及数据分析方式缺乏透明度。在本次讨论中，我建议对模型进行修改并简化分析。我对这两个案例研究进行了更详细的分析。该分析中获得的 P 值与论文中获得的 P 值接近。最后，提出了对多重测试的更清晰的修正，并对“仅 P 值”缺乏解释进行了评论。