Multiple-Attempt, Single-Item Response Models

Abstract

The psychomotor literature has seen the appearance of several papers that have described the proposed or actual use of item resonse theory (IRT) models in the measurement of psychomotor skills (Costa et al., 1989; Safrit et al., 1989; Spray, 1987, 1989). These IRT models were originally developed for use with the assessment of either cognitive or affective behaviors but could be used on psychomotor responses in some situations. In order to use these models with psychomotor responses, the data have been frequently treated as though the responses had been obtained from mental tests. For example, a study by Safrit et al. (1992) examined the number of sit-ups that an examinee could perform within a 60-second time limit The number of sit-ups completed were dichotomized as pass/fail or 1/0 responses according to a cutoff or criterion score, so that the data could be used in an IRT model commonly used for multiple-choice items. Several sit-up tests were administered to an examinee and each test varied by difficulty. This series of sit-up tests resulted in a response vector or string of dichotomous responses for each examinee, which were then treated as typical 1/0 or right/wrong responses to a cognitive test. The dichotomized responses were subsequently fitted to a two-parameter logistic model (2-PLM).