Gnambs Timo, Staufenbiel Thomas
Scaling and Test Design, Leibniz Institute for Educational Trajectories, Bamberg, Germany.
Institute of Psychology, Osnabrück University, Osnabrück, Germany.
Res Synth Methods. 2016 Jun;7(2):168-86. doi: 10.1002/jrsm.1190.
Two new methods for the meta-analysis of factor loadings are introduced and evaluated by Monte Carlo simulations. The direct method pools each factor loading individually, whereas the indirect method synthesizes correlation matrices reproduced from factor loadings. The results of the two simulations demonstrated that the accuracy of meta-analytical derived factor loadings is primarily affected by characteristics of the pooled factor structures (e.g., model error, communality) and to a lesser degree by the sample size of the primary studies and the number of included samples. The choice of the meta-analytical method had a minor impact. In general, the indirect method produced somewhat less biased estimates, particularly for small-sample studies. Thus, the indirect method presents a viable alternative for the meta-analysis of factor structures that could also address moderator hypotheses. Copyright © 2016 John Wiley & Sons, Ltd.
本文介绍了两种用于因子载荷元分析的新方法，并通过蒙特卡罗模拟对其进行了评估。直接法单独汇总每个因子载荷，而间接法则综合从因子载荷中重现的相关矩阵。两次模拟的结果表明，元分析得出的因子载荷的准确性主要受汇总因子结构的特征（如模型误差、共同度）影响，而受原始研究样本量和纳入样本数量的影响较小。元分析方法的选择影响较小。一般来说，间接法产生的偏差估计略少，尤其是对于小样本研究。因此，间接法为因子结构的元分析提供了一种可行的替代方法，该方法也可以处理调节变量假设。版权所有© 2016约翰威立父子有限公司。
