skip to main content skip to footer


Wingersky, Marilyn S.
Publication Year:
Report Number:
ETS Research Report
Document Type:
Page Count:
Subject/Key Words:
Computer Software, Item Parameter Estimation, Item Response Theory (IRT), LOGIST


Applications of item response theory frequently require estimates of only the item parameters. In this paper, a reparameterization of the logistic model is developed which involves substituting an implicit function of the number-right true score for ability. The number-right true score needs only to be computed once since the observed number-right score is a good estimator of it. This reparameterization avoids estimating each examinee's ability in every stage of the estimation procedure. An estimation method containing this reparameterization was devised and incorporated into a modified version of the LOGIST computer program. The procedure, called One-Stage, was tried out on artificial and real data. The results on the artificial data were compared to the true values and to the estimates from the operational LOGIST program. The results on the real data were compared to the estimates from the operational LOGIST program. In the comparisons of the parameter estimates from One-Stage and LOGIST to the true values for the artificial data, One-Stage did not estimate the item parameters as well as LOGIST. One potential benefit of this procedure, the reduction in computer costs due to not estimating an ability for each examinee in each stage, was not realized when the procedure was applied to real data with a lot of not reached items. (54pp.)

Read More