Key points are not available for this paper at this time.
Background: The Western Aphasia Battery (WAB) (Kertesz, 1982 Kertesz, A. 1982. Western Aphasia Battery, New York: Grune & Stratton. Google Scholar) is one of the most frequently used tests of general language performance in aphasia, despite significant psychometric limitations. Item response theory (IRT) provides measurement models that may address some of these limitations. Aims: The purposes of this investigation were to evaluate whether the WAB can be productively fit to an IRT model, and to evaluate whether IRT modelling confers psychometric benefits. Methods & Procedures: An analysis of WAB data collected from a convenience sample of 101 individuals with chronic aphasia was undertaken. Exploratory factor analysis was conducted to evaluate the dimensionality of the WAB, and a Rasch Partial Credit Model was fit to the data. Item fit statistics and residual correlations were used to test key model assumptions. The distribution of traditional and Rasch-based person scores, and the relationship between ability level and test reliability were also examined. Outcomes & Results: Despite reasonable overall fit to the model, a small number of WAB items demonstrated significant misfit, suggesting that they do not productively contribute to the measurement of aphasia severity. A small but substantial minority of individuals also demonstrated inadequate fit to the measurement model. Rasch-based scores were more normally distributed than traditional scores, and score reliability varied substantially across the ability range. Conclusions: These results suggest that the WAB may be productively fit to an IRT-based measurement model, and that such models may be used to improve the psychometric properties of aphasia tests. Benefits include indices of severity and score reliability that are more valid than those currently in use, and the potential for improved efficiency of testing through adaptive administration.
Hula et al. (Tue,) studied this question.