The impact of missing data on the dichotomous mixture IRT models
MetadataShow full item record
A mixture IRT model introduces latent classes into the model and is used when researchers want to model the heterogeneity of a population. It is designed to separate a population of individuals into qualitatively or quantitatively di erent subgroups or classes. Even though the mixture IRT model has turned out to be useful in many applications, the fundamental concern in using the mixture IRT model is whether the underlying population structure identi ed by the model really exists or not. The flexibility of the mixture IRT model may produce spurious classes to t data better. A number of extended mixture IRT models have been developed and applied to various applications in IRT. It is well known that missing data can occur in many contexts and can result in biased parameter estimates. This study investigated whether the missing data could bias parameter estimates and also could produce additional classes in the dichotomous mixture IRT models. A simulation study was conducted to investigate the impact of missing data on the bias, root mean square error (RMSE), and the rate of correct identi cation of classes in mixture IRT models. The results of simulation study showed that missing data could introduce bias in parameter estimation and spurious classes in class identi cation.