Explanatory Secondary Dimension Modeling of Latent Differential Item Functioning
Paul De Boeck, Sun-Joo Cho and Mark Wilson
The study works with latent classes （ a non-DIF latent class and a DIF latent class ）and takes a mixed dimensionality approach （ the secondary dimension plays a role in one the DIF latent class but not in the non-DIF latent class ） in order to model and e xpl ain DIF .
The One-Dimensional Mixture Models , the Mixture of One-Dimensional and Two-Dimensional Model and the T wo-Dimensional Mixture Models are introduced. A college-level mathematics placement test and an Arithmetic Operations test were used as the examples. The computer program LatentGOLD 4.5 Syntax module (Vermunt & Magidson, 2007) was used to estimate the model parameters. (AIC) and Bayesian information criterion (BIC) are computed for models’ comp ar ison. The results demonstrate that the mixture approach makes sense and that a secondary dimension leads to a better goodness of fit for the data sets under consideration .
Comments:
The thinking is always precious.
I do not figure out the latent group thing in this paper. It seems that first we should to divide the participants into two groups, then the mixed model will be used for explanation. In the paper, the author brought up that the focus of this article is to explain DIF rather than detect DIF. In my opinion, if there is no DIF, then there is no need to explain DIF, so first we must make sure the DIF is really existing，then we use his model to explain them. Whether the number of DIF will be different from the former methods? If there are differences then which one is correct? .