Parameter Estimation for Finite Mixtures of Generalized Partial Credit Models
Internaional Conference on Computer Systems and Technlologies – CompSysTech'2008, Gabrovo
We consider a finite length test and assume that each item has graded response categories which satisfy the Generalized Partial Credit Model (GPCM) from the Item Response Theory [1-4] and also that there exist population classes. We give a parameter estimation procedure for this finite GPCM mixture, based on the EM algorithm and test the model with simulated data. The paper contains complete (and formal alternative to ) derivation of the basic model formulas.
Item Response Theory, Generalized Partial Credit Model, EM algorithm
Природни науки, математика и информатика
Natural sciences, mathematics and informatics