Investigation of the effect of parameter estimation and classification accuracy in mixture IRT models under different conditions

This study aims to examine the effects of mixture item response theory (IRT) models on parameter estimation and classification accuracy under different conditions. The manipulated variables simulation are set as IRT (Rasch, 2PL, 3PL); sample size (600, 1000); number items (10, 30); latent classes (2, 3); missing data type (complete, at random (MAR) not (MNAR)), percentage (10%, 20%). Data were generated for each three using code written in R program. MplusAutomation package, which provides automation Mplus program, was used analyze data. mean RMSE values difficulty, discrimination, guessing determined. Mixture Rasch model found be lower than those 2PL 3PL models. Percentages also computed. It noted that with 30 items, 2 classes, 1000 size, complete conditions had highest percentage. Additionally, a factorial ANOVA evaluate factor's main interaction effects.