Marginal maximum likelihood estimation in polytomous Rasch models using SAS
Résumé
The growing use of Patient Reported Outcomes (PRO's) calls for user friendly software to fit IRT models in standard software like SAS or R. This paper describes a SAS macro %rasch_mml that fits polytomous Rasch models. The macro estimâtes item parameters using marginal maximum likelihood (MML) estimation and person locations using both MLE and Warm's Weighted likelihood estimation (WLE). A number of standard graphical présentations that are useful for investigating the properties of items are included: plots of item characteristic curves (ICC's), personitem location maps (Wright maps) comparing estimâtes of person locations and item locations, and item and test information functions. A graphical goodness-of-fittest is also produced.
Domaines
Statistiques [math.ST]
Origine : Accord explicite pour ce dépôt