TitleThe Rasch model and missing data, with an emphasis on tailoring test items
Publication TypeConference Paper
Year of Publication1988
Authorsde Gruijter, DNM
Conference Nameannual meeting of the American Educational Research Association
Date PublishedApril 5-9
Conference LocationNew Orleans, LA. USA
Accession NumberED297012
Abstract

Many applications of educational testing have a missing data aspect (MDA). This MDA is perhaps most pronounced in item banking, where each examinee responds to a different subtest of items from a large item pool and where both person and item parameter estimates are needed. The Rasch model is emphasized, and its non-parametric counterpart (the Mokken scale) is considered. The possibility of tailoring test items in combination with their estimation is discussed; however, most methods for the estimation of item parameters are inadequate under tailoring. Without special measures, only marginal maximum likelihood produces adequate item parameter estimates under item tailoring. Fischer's approximate minimum-chi-square method for estimation of item parameters for the Rasch model is discussed, which efficiently produces item parameters. (TJH)