How Do Different Weighting Methods Affect the Overall Effect Size in Meta-Analysis? : An Example of Science Attitude in Türkiye Sample

Abstract

There is increasing interest in meta-analysis in different fields due to the need to combine the results of primary research. One of the crucial concepts in combining results is weighting. This study examines how Hunter and Schmidt's method, weighting by sample size; Hedges and Vevea's method, weighting by inverse variance; and Osburn and Callender's method, unweighting, affect the overall effect size in meta-analysis. In this context, for meta-analysis, the search was done for studies examining the effects of alternative measurement and assessment techniques and methods in science education on science attitudes. The databases of the HEI National Thesis Center, Web of Science, ERIC, EBSCO, Google Scholar, and DergiPark were searched between 2010 and 2021. Eleven studies (with 14 effect sizes) that met the criteria were included in the meta-analysis. In line with the study's findings, it was observed that the overall effect sizes were significant and did not change much in the weighting methods. Besides, it was found that the method with the lowest standard error was unweighted. The weighting methods of Hunter and Schmidt and Hedges and Vevea gave similar results in terms of standard error. When the correlation coefficient between the weighting methods was examined, it was seen that all correlation coefficients were greater than 0.90.