When conducting meta-analysis, reviewers gather extensive sets of primary studies for meta-analysis. When we have two or more primary studies by the same author, or two more studies using the same data set, we have the issues we call ‘same author’ and ‘same data’ issues in meta-analysis. When a researcher conducts a meta-analysis, he or she first confronts ‘same author’ and ‘same data’ issues in the data gathering stage. These issues lead to between studies dependence in meta-analysis.
In this dissertation, methods of showing dependence are investigated, and the impact of ‘same author’ studies and ‘same data’ studies is investigated. The prevalence of these phenomena is outlined, and how meta-analysts have treated this issue until now is summarized. Also journal editors’ criteria are reviewed.
To show dependence of ‘same author’ studies and ‘same data’ studies, fixed-effects categorical analysis, homogeneity tests, and intra-class correlations are used. To measure the impact of ‘same author’ and ‘same data’ studies, sensitivity analysis and HLM analyses are conducted. Two example analyses are conducted using data sets from a class-size meta-analysis and ESL (English as a Second Language) meta-analysis. The former is an example of the ‘same data’ problem, and the latter is an example of the ‘same author’ problem. Finally, simulation studies are conducted to assess how each analysis technique works.
