Type I and Type II Error Rates and Overall Accuracy of the Revised Parallel Analysis Method for Determining the...

Area of Science:

• Psychometrics
• Statistical analysis
• Factor analysis

Background:

• Estimating the correct number of factors is crucial in factor analysis.
• Traditional parallel analysis (T-PA) and Revised parallel analysis (R-PA) are established methods.
• Sequential likelihood ratio test (LRT) methods are commonly used alternatives.

Purpose of the Study:

• To compare the accuracy of T-PA, R-PA, and LRT methods.
• To evaluate the performance of these methods across various conditions.

Main Methods:

• Monte Carlo simulation was employed to assess method accuracy.
• T-PA and R-PA were conceptualized as stepwise hypothesis-testing procedures.
• Comparison involved sequential eigenvalue comparisons.

Main Results:

• Parallel analysis approaches generally outperformed LRT methods.
• R-PA showed improved accuracy compared to T-PA, particularly with higher factor loadings.
• No single method was uniformly superior across all simulated conditions.

Conclusions:

• PA methods, especially R-PA, offer a more accurate approach to determining the number of factors.
• R-PA's enhanced performance in specific conditions suggests its utility in complex factor structures.
• Findings support the use of PA over LRT for factor determination in many statistical applications.