Robustness of marginal maximum likelihood estimation in the Rasch model
Simulation studies examined the effect of misspecificationof the latent ability (θ) distribution on the accuracyand efficiency of marginal maximum likelihood(MML) item parameter estimates and on MML statisticsto test sufficiency and conditional independence. Resultswere compared to the conditional maximum likelihood(CML) approach. Results showed that if θ is assumedto be normally distributed when its distributionis actually skewed, MML estimators lose accuracy andefficiency when compared to CML estimators. The effectsare not large, though they increase as the skewnessof the number-correct score distribution increases.However, statistics to test the sufficiency and conditionalindependence assumptions of the Rasch modelin the MML approach are very sensitive to misspecificationof the θ distribution. Index terms: ability distribution,conditional likelihood, efficiency, goodnessof fit, marginal likelihood, Rasch model, robustness.
Year of publication: |
1990
|
---|---|
Authors: | Zwinderman, Aeilko H. ; Van den Wollenberg, Arnold L. |
Saved in:
freely available
Saved in favorites
Similar items by person
-
Consistency of Rasch model parameter estimation: A simulation study
Van den Wollenberg, Arnold L., (1988)
-
A joint latent class changepoint model to improve the prediction of time to graft failure
Garre, Francisca Galindo, (2008)
- More ...