Comparison of some methods of estimating Rasch model parameters

Abstract

An estimation procedure for the Rasch model parameters was described by performing three methods (the greatest possibility estimation method, the Cohen approximation method, and the Bayesian estimation method), and comparing them with the simulation method, for several sample sizes (the number of individuals, the number of items) and for several fixed values ​​for the parameters of the distributions used in the simulation. And from the comparison of the methods to find the estimate of the item difficulty parameter (), and the individual ability parameter (), the best value for the parameters of the Rasch model for the discrete and continuous distributions was the estimate by the Bayesian method, based on the standard of the mean absolute relative error (MAPE).