Abstract
Differential item functioning in heterogeneus populations. Latent classes. The existence of qualitatively different latent classes within a population demand the adaptation of traditional procedures to detect DIF. The aim of this study was to explore the relations between DIF and latent classes. In order to achieve this, a numerical aptitude test was analyzed. The results show a decrease of the number of items with DIF; 28% in the traditional way versus 16% if the mixed Rasch model is used before DIF detection. In DIF analysis studies it would be more appropriate to estimate the latent classes and then to define the reference/focal groups within these classes.
