The Mixture Response Time Model For Detecting Unexpected Rapid-guessing Behaviors

Abstract

Rapid guessing, which means that respondents tend to respond to an item quickly by guessing the answers, is commonly observed in ability tests. In particular, rapid guessing is not a consistent status for a respondent; instead, a respondent can rapidly guess on any item due to non‐motivation, speededness, responding strategy, or whatever reasons. By the aids of the development of computer‐based testing, timing information can be a supportive evidence for identifying if a respondent endorses a response thoughtfully or dismissively. In this study, a mixture item response model incorporating response time is proposed to detect unexpected rapid guessing behaviors during testing. We conducted a simulation study to evaluate the parameter recovery of the new model by using the Marcov chain Monte Carlo estimation with Gibbs sampling. Results showed that the parameters in the new model can be recovered fairly well; ignoring rapid‐guessing behaviors by fitting a standard item response model would overestimate item difficulties and underestimate item item‐intensities seriously. Finally, a computer-based mathematical test was analyzed as an example for demonstration.