Enhancing Factor Analysis: Evaluating Factor Retention Accuracy With Measurement Error

• In Event: Factor Analysis and Structural Equation Modeling in Educational Research

Abstract

Exploratory factor analysis(EFA) employs various criteria to identify the correct number of factors in data. This study investigated the performance of extreme gradient boosting(xgboost), a machine learning algorithm, in comparison to traditional factor retention criteria such as Kaiser criteria, parallel analysis, and empirical Kaiser criterion. Using simulated data with varying conditions, including different sample sizes, factor correlations, factor numbers, and variable numbers per factor, we evaluated the accuracy of each method. Results revealed that xgboost consistently outperformed other criteria in determining the number of factors, even when incorporating variables without factor associations. Additionally, the presence of non-associated variables significantly impacts the performance of the Kaiser criteria. These findings provide valuable insights for researchers utilizing EFA in diverse settings.

Authors

• Youngwon Kim, Harvard University

• Elizabeth A. Sanders, University of Washington