EVIDENCE ON THE SIMPLE STRUCTURE AND FACTOR INVARIANCE ACHIEVED BY FIVE ROTATIONAL METHODS ON FOUR TYPES OF DATA.

Area of Science:

• Psychometrics
• Statistical analysis
• Data mining

Background:

• Factor rotation is crucial for interpreting principal components analysis (PCA) and factor analysis (FA).
• Evaluating different rotation methods is essential for optimizing data interpretation and ensuring reliable results.
• Simple structure and factor invariance are key criteria for assessing rotation method performance.

Purpose of the Study:

• To compare the effectiveness of five factor rotation methods: Maxplane, Oblimax, Promax, Harris-Kaiser, and Varimax.
• To assess these methods across diverse data types: questionnaire, objective test, physical problem, and plasmode.
• To evaluate rotation performance based on simple structure (hyperplane percentages) and factor invariance (congruence coefficient).

Main Methods:

• Application of Maxplane, Oblimax, Promax, Harris-Kaiser, and Varimax to four distinct datasets.
• Utilizing Rotoplot-assisted visual rotations following the Maxplane procedure.
• Comparative analysis using hyperplane percentages for simple structure and congruence coefficients for factor invariance.

Main Results:

• Oblique methods generally outperformed Varimax in achieving simple structure.
• Factor invariance results were not consistently superior for oblique methods compared to Varimax.
• Rotoplot-assisted Maxplane often maximized simple structure at a .10 hyperplane width.
• Unassisted Maxplane was outperformed by less computationally intensive oblique methods in both criteria.
• The Harris-Kaiser method demonstrated a favorable balance between simple structure and factor invariance.

Conclusions:

• Oblique rotation methods are generally preferable to Varimax for enhancing simple structure in factor analysis.
• The Harris-Kaiser method presents a robust and satisfactory option when considering both simple structure and factor invariance.
• The choice of rotation method significantly impacts the interpretability and invariance of factor solutions across different data types.