Investigation of the Mediator Variable Effect Using BK, Sobel and Bootstrap Methods (Mathematical Literacy Case)

Abstract

This study aimed to compare different mediation analysis methods (BK, Sobel, and bootstrapping) based on single mediation models for groups of different sizes. For this purpose, the PISA 2012 data for Turkey were used. In order to compare the mediation analysis methods, 4,848 students from Turkey that participated in PISA 2012 were divided into sample groups of 100, 200, 500 and 1,000 individuals. Among the mediation analysis methods discussed within the scope of the research, the BK method was implemented assisted by a regression analysis while for the remaining two methods, SPSS macros were utilized. For the analysis, syntax files were created to be run on SPSS. The results of the analysis of single mediation models revealed that the mathematics anxiety variable mediated the relationship between classroom climate and mathematical literacy. According to the analyses based on all three methods, it was observed that the standard error value increased as the sample group became smaller. Although the standard errors of the Sobel test and bootstrap method were close to each other in large study groups, the former produced less erroneous results in large samples whereas the latter yielded more reliable results in smaller samples.