AUTHOR : 황성동
INFORMATION : page. 1~19 / 2016 Vol.23 No.1
As a new mode of research synthesis, meta-analysis has recently drawn much attention from researchers, having proved to produce less biased results than the traditional research synthesis method. However, scholars may be suspicious about the results of meta-analysis due to publication bias as it occurs when published studies represent a biased selection of the evidence. This study was designed to discuss the meaning of publication bias and conduct publication bias analysis using an example data set so as to provide researchers with a guideline for performing valid meta-analysis. The process of publication bias analysis begins with a careful examination of the effect sizes of individual studies and continues with testing funnel plot asymmetry. With both a graphical test and a statistical one, potential publication bias could not be ruled out. To deal with this bias, adjustments were made using the trim-and-fill method followed by the Copas selection model, and the original result was found to be valid and statistically robust even with the small bias detected. In the academic circle of meta-analysis, publication bias analysis is specified as one of the required items in some guidelines of reporting and evaluating the results of meta-analysis. Therefore, it is highly recommended to conduct publication bias analysis for the validity of meta-analysis as well as the integrity of individual meta-analytic studies.