Coefficient Alpha and Interculture Test Selection

Abstract

The internal consistency reliability of a measure can be a focal point in an evaluation of the potential adequacy of an instrument for adaptation to another cultural setting. Cronbach’s alpha (α) coefficient is often used as the statistical index for such a determination. However, alpha presumes a tau-equivalent test and may constitute an inaccurate population estimate for multidimensional tests. These notions are expanded and examined with a Japanese version of a questionnaire on nursing attitudes toward suicidal patients, originally constructed in Sweden using the English language. The English measure was reported to have acceptable internal consistency (α) albeit the dimensionality of the questionnaire was not addressed. The Japanese scale was found to lack tau-equivalence. An alternative to alpha, “composite reliability,” was computed and found to be below acceptable standards in magnitude and precision. Implications for research application of the Japanese instrument are discussed.