A Simulation Study Of Three Methods For Determining The Number Of Image Components

Area of Science:

• Psychometrics
• Statistical analysis
• Data mining

Background:

• Component extraction is crucial for data analysis.
• Selecting appropriate stopping rules impacts results.
• Existing rules vary in performance.

Purpose of the Study:

• Compare the accuracy of three stopping rules for image component extraction.
• Identify the most robust rule under varying conditions.

Main Methods:

• Simulation-based experimental design.
• Generated simulated correlation matrices.
• Varied sample size, number of variables, loading magnitudes, and component correlations.

Main Results:

• All rules performed accurately under favorable conditions (high sample size, high loadings).
• Veldman's rule, utilizing the varimax criterion, demonstrated superior accuracy in challenging conditions.
• Performance varied significantly with changes in experimental parameters.

Conclusions:

• Veldman's rule is recommended for image component extraction, especially in complex datasets.
• The choice of stopping rule is critical and depends on data characteristics.
• Further research should explore additional stopping rules and criteria.