Evaluating cluster analysis solutions: an application to the Italian NEO personality inventory

This paper is concerned with the evaluation of cluster analysis solutions. Internal criteria and replication issues are compared and applied to empirical data collected from an Italian sample of 421 young adults, using the NEO Personality Inventory. The following internal criteria were considered: C, gamma, and G( + ) indices, and point‐biserial correlation. Replication was examined (i) ‘internally’ using double cross‐validation and bootstrap approaches and (ii) ‘externally’ by comparing the solution obtained on the Italian sample with the results obtained in German and Spanish samples. While replication analyses supported three‐ and four‐cluster solutions, internal criteria (with the exception of point‐biserial correlation) tended to privilege solutions with a much larger number of groups. Advantages and limitations of the different strategies are discussed. Copyright © 2002 John Wiley & Sons, Ltd.