The biasing properties of Cronbach alpha in fixed-format tests were examined in a Monte Carlo simulation under conditions of varying shapes of data distribution. The distributions of data examined were normal (3, 1), uniform (1, 5), skewness biased, Kurtosis biased (leptokurtic), and U-shaped. Analysis indicated that alpha was significantly different from zero under the normal and uniform distributions; however, negative interitem correlations and marginal SDs might have been responsible for that effect. The magnitude of alpha was negligible so it is suggested that researchers can use alpha with data having various distributional properties.
Get full access to this article
View all access options for this article.
References
1.
BardoJ. W.HugheyJ. B. (1978) A Monte Carlo simulation of effects of central tendency error on the magnitude of Cronbach's coefficient alpha. Perceptual and Motor Skills, 46, 308–310.
2.
CronbachL. (1951) Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297–334.
3.
MatyasT. A.GreenwoodK. M. (1991) Problems in the estimation of autocorrelation in brief time series and some implications for behavioral data. Behavioral Assessment, 13, 137–157.
4.
NunnallyJ. C. (1967) Psychometric theory. New York: McGraw-Hill.
