15.3 IRT Models
IRT models can be classified as a function of the trait being measured (ability versus personality/person-center), the response format (dichotomous versus polytomous items), and the number of parameters (1P versus 2P versus 3P...). Table 15.3 lists the most popular unidimensional IRT models.
| Dichotomous | Polytomous |
|---|---|
| Rasch model (1P) | Partial credit model (PCM) |
| 1-Parameter logistic model (1PLM) | Rating scale model (RSM) |
| 2-Parameter logistic model (2PLM) | Generalized partial credit model (GPCM) |
| 3-Parameter logistic model (3PLM) | Graded response model (GRM) |
| 4-Parameter logistic model (4PLM) | Nominal response model (NRM) |
It is beyond the scope of this chapter to cover non-parametric models (e.g., Mokken), 4PL or 5PL models (e.g., unfolding), complex polytomous models (e.g., GRM, NRM), or IRT multidimensional models for simple or bifactor structures (Ostini & Nering, 2006; Paek & Cole, 2020; Revelle, 2023). For this reason, only some of the most popular IRT models (i.e., 1PLM, 2PLM, 3PLM) will be briefly described in the following subsection.