Guessing and cognitive diagnostics: A general multicomponent latent trait model for diagnosis
Lutz, Megan Elyse
MetadataShow full item record
A common issue noted by detractors of the traditional scoring of Multiple Choice (MC) tests is the confounding of guessing or other false positives with partial knowledge and full knowledge. The current study provides a review of classical test theory (CTT) approaches to handling guessing and partial knowledge. When those methods are rejected, the item response theory (IRT) and cognitive diagnostic modeling (CDM) approaches, and their relative strengths and weaknesses, are considered. Finally, a generalization of the Multicomponent Latent Trait Model for Diagnosis (MLTM-D; Embretson & Yang, 2013) is proposed. The results of a simulation study are presented, which indicate that, in the presence of guessing, the proposed model has more reliable and accurate item parameter estimates than the MLTM-D, generally yielding better recovery of person parameters. Discussion of the methods and findings, as well as some suggested directions for further study, is included.