A Monte Carlo study of the validity of the Inverse Scree test under conditions where true group membership is known was presented. Four distributions involving 2, 3, 4, and 5 true groups of 3,000 computer "subjects" were generated using nonredundant measurement on four orthogonal variables. A fifth Low-N distribution for 3 true groups was also generated. Fifty cluster analyses of each distribution were analyzed both by the JE(2)/JE(l) mathematical criterion and by four raters using the Inverse Scree technique. The results indicated that JE(2)/JE(l) is markedly N-dependent and that Inverse Scree tests are affected by large numbers of true groups. Implications for the data analyst are discussed.
Get full access to this article
View all access options for this article.
References
1.
Cattell, R. B. (1978). The scientific use of factor analysis. New York: Plenum.
2.
Dillon, W. R. and Goldstein, M. (1984). Multivariate analysis: Methods and applications. New York: Wiley.
3.
Duda, R. O. and Hart, P. E. (1973). Pattern classification and scene analysis. New York: Wiley.
4.
Lathrop, R. G. and Williams, J. E. (1987). The reliability of inverse scree tests for cluster analysis. Educational and Psychological Measurement, 47, 953-959.
5.
Lathrop, R. G. and Williams, J. E. (1989). The shape of the inverse scree test for cluster analysis. Educational and Psychological Measurement, 49, 767-780.
6.
Milligan, G. W. and Cooper, M. C. (1985). An examination of procedures for determining the number of clusters in a data set. Psychometrika, 50, 159-179.
