A Monte Carlo PCA Approach to Parallel Analysis is an objective, simulation-based statistical technique used to determine the exact number of components or factors to retain in Principal Component Analysis (PCA) and Exploratory Factor Analysis (EFA). Originally developed by John L. Horn in 1965, this method addresses a major flaw in traditional retention rules: the tendency to overextract and mistake random noise for true statistical patterns. The Core Problem It Solves
Traditional extraction criteria often fail because they do not account for sampling error or least-squares capitalization:
The Kaiser Criterion (Eigenvalue > 1) regularly overestimates the number of significant factors. Even with completely uncorrelated variables, a sample PCA will naturally yield some eigenvalues greater than 1 purely due to random chance.
The Scree Plot relies on a subjective visual assessment of where the curve “bends” or “elbows,” leaving room for researcher bias. How the Monte Carlo Simulation Works
Parallel Analysis replaces subjectivity with empirical simulation. The process follows these automated steps:
Leave a Reply