Randomly Clicking on Experimental Items and Item Parameter Estimation
Xiaoliang Zhou · Proceedings of the 2019 AERA Annual Meeting · 2019
The present study aims to explore how giving up on experimental items and randomly clicking on them may affect item parameter estimation in IRT models.We examine the extent to which randomly clicking on experimental items may impact the item parameter estimation of 2PL models.Using a simulation study, it was found that increasing the proportion of randomly clicking examinees increased the bias and RMSE for both discrimination and difficulty parameter estimations.In addition, increasing the number of experimental items reduced the bias and RMSE of both the discrimination parameters and difficulty parameters.Finally, parameter recovery for both discrimination and difficulty, in terms of bias and RMSE, was the most robust to random clicking when the population's ability was around 0 or standard deviation was smaller, except that the bias of difficulty was more serious for smaller standard deviations.