The Analysis of Partial Credit Scoring

Abstract

This article discusses a range of issues in the practical application of an item response theory (IRT) method for partial credit scoring. After a brief discussion of partial credit scoring as an alternative to right-wrong scoring in the measurement of educational achievement, an IRT model for partial credit analysis is developed and described. This model is presented as a straightforward and logical application of Rasch's dichotomous model to a sequence of ordered response alternatives. The distinctive nature of the item parameters in the model is described and these parameters are contrasted with two more familiar sets of parameters: Thurstone thresholds and the difficulties of dichotomously scored subitems. Issues in marking out and interpreting variables using this model are discussed. Brief mention is made of several special cases of the partial credit model that may be useful in particular applications and for particular kinds of test and questionnaire data.