Electronic Theses and Dissertations



Document Type


Degree Name

Doctor of Philosophy





Committee Chair

Kristoffer Berlrin

Committee Member

Randy Floyd

Committee Member

George Relyea

Committee Member

Stephanie Huette


Many considerations influence the results of exploratory factor analysis (EFA) including rotation criteria and estimation method. Past research has given recommendations for matching estimation method to measurement of indicator variables, and some recommendations exist for the choice of rotation criteria across item complexity and factor intercorrelation. Less guidance exists, however, on rotation recommendations across number of factors and indicators included in the model. To fill this gap, the present study manipulated the number of factors and indicators in EFA models with ordinal data across multiple rotation criteria and across additional data considerations. Using Monte Carlo methods, data were generated (1,000 replications per condition) in Mplus 8.3 manipulating number of factors (two-factor and three-factor models) and number of indicators (five-, ten-, and fifteen-indicators per factor). Data were analyzed across rotation criteria (Geomin, Quartimin, CF-Equamax, and CF-Facparsim) and additional data manipulations included sample size, factor structure, and amount of intercorrelation between factors. Chi-square tests and ANOVAs were conducted to investigate differences in model convergence and model results, respectively, considering both effect sizes and practical differences between conditions. No practical differences were observed across number of indicators for convergence, relative bias, or congruence results. For number of factors, however, an increase from two- to three-factors had a practical effect, especially on congruence results as item complexity and factor intercorrelation increased. Rotation results indicate that choice of rotation criteria becomes more important with an increase in factors. In conditions of complex structure, especially at larger levels of intercorrelation, CF rotations produce more stable results with the addition of a factor. Because complexity and intercorrelation cannot be known a priori, it is recommended to conduct and report EFA across rotations which minimize both row and column complexity.


Data is provided by the student.

Library Comment

dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.