Item-level response times increasingly play an important role in educational and psychological testing, and as such can be adversely affected by undesirable response behavior, such as speeding and innatentive responding.
Test speededness, for example, can cause response times to deviate from the hypothesized model. In this article, we introduce a robust estimation approach when estimating a respondent’s working speed under the log-normal model by down-weighting aberrant response times.
A simulation study is carried out to compare the performance of two weighting schemes and a real data example is provided to showcase the use of the new robust estimation method.
Hong, M., Rebouças, D. A., & Cheng, Y. (accepted). Robust Estimation for Response Time Modeling. Journal of Educational Measurement.