Minorization Algorithms For The Rasch Model
Door: Maris, G.K.J., Bechger, T.M. | 01-01-2009 From the principle of minorization, we derive monotone convergent algorithms for conditional and marginal maximum. likelihood estimation in the Rasch model, where the parameters are updated one item at the time.In addition, we show that these algorithms can be made faster by deliberately over-parameterizing the model. The algorithm for CML turns out to be equal to the implicit equations algorithm that was proposed by Gerhard Fischer in the early 1970s, without proof of its monotone convergence.
In this paper we consider estimation methods for the parameters in the Rasch (1960) model based on the idea of minorization (De Leeuw, 1994). Minorization provides a general framework for constructing monotone convergent algorithms_ for parameter estimation.
Read more
Kunnen we je helpen?
Stel je vraag via onze kanalen of kijk in de veelgestelde vragen.
Voor scholen: Vergeet niet om het brinnummer bij de hand te hebben en/of in de mail te vermelden, zodat we jouw vraag sneller kunnen behandelen!