Meta-Analysis within a Single Paper Using Individual-Participant Data: A Simpler Alternative

Abstract

More researchers recognize the importance of meta-analyzing findings within a single paper (i.e., mini meta-analysis). These analyses have been primarily conducted by meta-analyzing aggregated data at the study level, particularly the effect size of each study. We highlight an alternative way of achieving the same goal by meta-analyzing data at the individual-participant level, which combines all the raw data into one dataset. We outline the procedure of conducting mini meta- analysis using the individual-level approach. In a series of 21 simulations, we compared the results between individual-level and the study-level meta-analyses in terms of the accuracy of effect size estimates, Type I errors, and Type II errors. We found that although the results are not identical, they are generally consistent, with the overall outcomes of mini meta-analyses at the individual-level being at least as good as those conducted at the study-level. We recommend more mini meta-analyses at the individual-level not only because individual-level meta-analysis has been regarded as the gold standard of meta-analysis, but also that it is a simple and convenient way to achieve the same goal.