Heterogeneity and negative skew often result in an underestimate of internal reliability when the usual split-half or alpha coefficients are computed for tests with a small number of items. A maximized correlation, although it may be positively biased, appears to be a more rational alternative. It was shown that r max is a stable statistic as well as a higher estimate of test reliability than alpha or r1l. Its use as a standard procedure was recommended.
Get full access to this article
View all access options for this article.
References
1.
Cronbach, L. J. (1943). On estimates of test reliability. The Journal of Educational Psychology, 34, 485-494.
2.
Feldt, L. S. , Woodruff, D. J., and Salih, F. A. (1987). Statistical inference for coefficient alpha. Applied Psychological Measurement, 11, 93-103.
3.
Guttman, L. (1945). A basis for analyzing test-retest reliability. Psychometrika, 10, 255-282.
4.
Kendall, M. , and Stewart, A. (1977). The advanced theory of statistics. (Vol. 1). New York: MacMillan Publishers.
5.
Wagner, E. E. , Alexander, R. A., Roos, G., and Prospero, K. (1985). Maximizing split-half reliability estimates for projective techniques. Journal of Personality Assessment, 49, 579-581.
6.
Wagner, E. E. , Alexander, R. A., Roos, G., and Adair, H. (1987). Optimum split-half reliabilities for the Rorschach: Projectives are more reliable than we think. Journal of Personality Assessment, 50, 107-112.
7.
Wagner, E. E. , Greene, R. S., Adair, H. E., and Dawson, J. (1987). Maximized reliability coefficients for LNNB factor scales. Journal of Personality Assessment, 9, 132-135.
8.
Wagner, E. E. and Flamos, O. V. (1988). Optimized split-half reliability for the Bender Visual Motor Gestalt Test: Further evidence for the use of maximization procedure. Journal of Personality Assessment, 52, 454-458.
9.
Woodruff, D. J. and Feldt, L. S. (1986). Tests for equality of several alpha coefficients when their sample estimates are dependent. Psychometrika, 51, 393-413.
