NCJ Number
88426
Date Published
Unknown
Length
30 pages
Annotation
Emphasizing European developments, this paper discusses two estimation procedures for the Rasch Model -- Unconditional Maximum Likelihood (UML) and Conditional Maximum Likelihood (CML) -- particularly with respect to new developments that make the more statistically rigorous CML estimation practical for use with 'longish' tests (those with more than 30 or 40 items).
Abstract
The Rasch Model is a talent trait model that assumes that the probability of a correct answer to an item is a function of the difficulty of that item and the ability of the person. Most researchers have been forced to use the unconditional procedures when applying the Rasch model to moderate-to-long tests. An algorithm has now been developed which makes the conditional procedures a feasible alternative. The conditional procedure is advantageous because of the known asymptotic properties of the estimates, which allows the use of goodness-of-fit tests. Another advantage of this procedure is the availability of the Andersen and Madsen method for estimating the parameters of the latent population distribution, and for testing hypotheses about this distribution. Equations and about 30 references are included.