Psychometric Strategies of Measures of Temperament: A Methodological Critique

Publisher Website
Google Scholar
Abstract
The test construction strategies used in the development of a large number of current questionnaire measures of temperament were reviewed. The review detailed the procedures that various researchers selected to construct their measures, including the specification of the content domain of interest, the method and sequence of item selection and evaluation, and efforts to determine the dimensionality of respective measures. A critique of these temperament measures was then provided, with reference to several central concepts in psychometric theory and multivariate developmental correlational research. The selection by some test constructors of coefficient alpha as the sole criterion for inferring unidimensionality was shown to be insufficient. Differences were discussed regarding the usefulness of the principal components and common factor models to represent multivariate item response data. Conceptual clarity and empirical inter-inventory studies were suggested to assist in the identification of similarity/dissimilarity among temperament constructs proposed by different investigators. A final recommendation was that greater attention be given to cross-validation as a component in the construction process of temperament measures.