每周论坛之三（总第218期）：柯紫筠（中山大学）：2019年9月16日下午

每周论坛之三  （总第218期）

报告人：柯紫筠博士

报告人单位：中山大学心理学系

题目：Correcting Publication Bias: New Methods and Software Implementation

时间：2019年 9 月 16 日 （周一，14:20-16:00）

地点：中山大学东校园心理学系（南学院楼C座）305

报告简介：

    Meta-analysis is increasingly recognized as a way to build and test theories. It has been widely used in fields like medical research, public health, social sciences, economics, management, etc. However, conclusions from meta-analyses may be misleading. Publication bias is one of the key factors that often lead to detrimental overestimation of the effects in meta-analyses. Recently, methodological researchers have developed a variety of new methods for correcting for publication bias. The goal of this presentation is to provide an intuitive overview of four classic and new methods, and to facilitate the implementations of those methods. In this presentation, we review the traditional “Trim-and-Fill” method, the meta-regression method, the p-curve and p-uniform methods, and the selection model methods. These methods are illustrated through a meta-analysis example on the effect of teacher expectancy. The general conclusion is that no single method consistently outperforms the other methods. The performance of any method in addressing the issue depends on whether its assumptions are satisfied. Except for small samples (i.e., the number of primary studies is small), the selection model methods are usually a better choice compared to the other methods. To facilitate the implementations of the newly developed methods, annotated R code is included. We end the presentation with a discussion on the challenges for future studies on publication bias correction.

报告人简介：

    柯紫筠，女，中山大学心理学系讲师。研究兴趣为元分析、结构方程模型元分析、增长曲线模型分析、时间序列分析、稳健分析等。