PAM: A double-precision FORTRAN routine for the parallel analysis method in principal components analysis

Abstract

Parallel analysis (Horn, 1965) is a Monte Carlo method of establishing the number of components in principal components analysis, and suggests retention of only those components from observed data whose eigenvalues are larger than the corresponding eigenvalues from random data. It can be seen as a correction of the eigenvalue-one rule for sample, rather than population, characteristics. PAM is a mainframe computer program that implements the parallel analysis method by providing mean and upper­ percentile eigenvalues from multiple replications of ran­ dom data. Both Monte Carlo studies (e.g., Zwick & Velicer, 1986) and studies with established data sets (e.g., Hubbard & Allen 1987) have found that parallel analysis yields ac­ curat~ results. Unfortunately, parallel analysis is not widely used, because at least one replication of random data must be generated. Typically, eigenvalues from sev­ eral replications of random data are averaged for e~ch or­ dinal position to provide criterion eigenvalues, WhICh are then compared to observed eigenvalues. .