Do Diagnostic Models Hold More Promise Than They Deliver?

• Activity Number: 399
• Type: Topic Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Matthias von Davier*+
• Companies: ETS
• Keywords: Latent class analysis ; Diagnostic classification models ; Categorical data analysis ; Discrete latent variables

Abstract

This talk will discuss recently developed models for complex student response data utilizing multidimensional discrete latent variables. These variables represent latent mastery/non-mastery levels which have to be inferred from observed responses to a series of items. Models of this type have been developed as extensions of latent class analysis to multiple classifications, and are nowadays often referred to as models for cognitive diagnosis, or diagnostic classification models. The class of models presented here is suitable for binary (correct/incorrect) and ordinal (partial credit) responses and allows modeling the latent structure as a combination of quantitative and qualitative latent variables. Applications of this general diagnostic modeling framework to analysis of test and questionnaire data from areas such as cognitive skills assessments, large scale surveys, longitudinal analysis, and the analysis of multidimensional personality tests (Big 5) will be discussed. The presentation will give an overview of the field, and will talk about promised advantages of using these models as well as limitations of these models.