Inferences from Bootstrap Method for Ability Parameters in 2-Parameter Logistic Model

Abstract

The ability parameter of persons/examinees estimates can be obtained using the Joint Maximum Likelihood (JML) estimation in Item Response Theory (IRT). However, JML estimates can be biased in some cases. Although the Bootstrap method has been considered for JML, existing studies remain far from satisfactory concerning the ability parameter estimation. This research evaluates the performances of JML and Bootstrap estimates of the ability parameter in terms of Standard Error Measurement (SEM) in the 2-Parameter Logistic (2-PL) model conducting a detailed Monte Carlo simulation study. According to the results, the average SEM estimates of the Bootstrap method are less than the average SEM estimates of JML in terms of the ability parameter.