B1306
Title: A new framework for jointly modelling response times and accuracy in computer-based learning tests
Authors: Maria Iannario - University of Naples Federico II (Italy) [presenting]
Abstract: A general modelling framework of response accuracy and response time is proposed to track skill acquisition and provide further diagnostic information on latent speed change in a learning environment. The contribution is placed in the area of structural equation models to jointly consider item responses designed using an ordinal level of accuracy, response times and item placement. Specifically, the hypothesis suggests that the response process is driven by two underlying latent variables: the latent variable representing the individual's ability as measured by the test items and the latent variable referring to the individual's speed in responding to the test items. The statistical model assumes that the item responses are directly influenced by the ability, while response times depend on both ability and speed. Consequently, response accuracy increases with the individual's skill level, while response time decreases with speed. A structure with increasing levels of complexity by relaxing (or not) certain parameter constraints is proposed for the analysis of a case study and a series of simulation experiments.