Abstract
In exploratory research, common factor analysis is often used for initial factor model specification. The model is then more rigorously analyzed using confirmatory factor analysis. The performance of several alternative, cluster-analytic approaches to initial model specification is evaluated here using population parameter analyses and a Monte Carlo simulation. Of the 6 cluster approaches evaluated, the one using the correlations of item correlations as a proximity metric and average linkage as a clustering algorithm (CCAL) performed the best. CCAL generally performed about as well or better than parallel analysis, but not as well as the maximum likelihood chi-square test in assessing dimensionality. The performance of CCAL is dramatically reduced in the presence of several cross-loadings. The use of CCAL in determining factor structure is further demonstrated using Hoyle and Lennox's (1991) self-monitoring data set.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have