The recent increase in interest for instructional relevance and fine-grained feedback from assessments has led to a unified paradigm of educational measurement, combining cognitive psychology with psychometrics, and thus, cognitive diagnostic assessment or CDA. CDAs are particularly useful for identifying areas of students' needs as well as designing individualized instruction and learning/teaching interventions to meet those needs. However, the typical CDAs assess coarsely defined attributes and lack information on the cognitive processes that underlie test performance.Cognitive processing takes time. A typical CDA is time-limited and the time an examinee allocates to tasks can provide insight on the cognitive process underlying the response. Response time (RT) has therefore been identified as important collateral information that can be used to account for examinee behavior in cognitive assessment. However, the use of RT in measurement models has, so far, been limited to approaches with the strict assumption that a test taker maintains a constant speed over the test process. In addition, most cognitive diagnostic modeling approaches have been directed towards classification of examinees based on their profiles on dichotomized status on the latent skill. Classifying latent attribute status into mastery and non-mastery not only obscures information but also ignores the fact that learning can be progressive, and respondents in the same category (mastery/non-mastery) may possess the skill to a considerably varying degree. These two concerns are the focus of the current study.This study aims to develop a more adaptable and informative modeling approach for examining and accounting for the effect of time speededness on examinees' cognitive processing behavior and ability in diagnostic models with polytomous attributes, thereby increasing the diagnostic potentials of CDAs. This is achieved by integrating variable working speed and partial mastery (polytomous attributes) into cognitive assessment model. The strengths of the model are assessed and compared to existing models using an empirical data and a simulation study. This new model, where applicable, allows for finer-grained feedback and flexibility in the assumed role of RT in cognitive diagnostic assessment while providing useful supplementary information to better understand testing strategies and behaviors.
Read
