Grade of Membership Response Time Model for Detecting Guessing Behaviors. (June 2016)

• Record Type: Journal Article
• Authors: Pokropek, Artur
• Is Part Of: Journal of educational and behavioral statistics. Volume 41:Number 3(2016)
• Page Start: 300
• Page End: 325
• Publication Date: 2016-06
• Subjects: guessing -- item filtering -- mixture models -- partial membership models -- response time; Educational statistics -- Periodicals; Social sciences -- Statistical methods -- Periodicals; 370.2
• DOI: 10.3102/1076998616636618
• Languages: English
• ISSNs: 1076-9986

Abstract

A response model that is able to detect guessing behaviors and produce unbiased estimates in low-stake conditions using timing information is proposed. The model is a special case of the grade of membership model in which responses are modeled as partial members of a class that is affected by motivation and a class that responds only according to the level of ability. Monte Carlo simulations were conducted to compare the proposed model with an approach that ignored guessing and an approach based on item filtering. In each simulated condition, the proposed model outperformed the other approaches by showing the lowest level of bias and the highest precision of item and persons estimates. Finally, the model was estimated using real life data from Programme for the International Assessment of Adult Competencies research (PIAAC). The results showed slight but expected corrections for the levels of proficiency in all countries.