An Extension of Least Squares Estimation of IRT Linking Coefficients for the Graded Response Model

Autor/in	Kim, Seonghoon
Titel	An Extension of Least Squares Estimation of IRT Linking Coefficients for the Graded Response Model
Quelle	In: Applied Psychological Measurement, 34 (2010) 7, S.505-520 (16 Seiten) PDF als Volltext Verfügbarkeit
Sprache	englisch
Dokumenttyp	gedruckt; online; Zeitschriftenaufsatz
ISSN	0146-6216
DOI	10.1177/0146621609344847
Schlagwörter	Least Squares Statistics; Computation; Item Response Theory; Models; Error of Measurement; Statistical Bias; Sample Size; Comparative Analysis + Suchen Sie Ihr Suchwort? Item-Response-Theorie; Analogiemodell; Messfehler
Abstract	The three types (generalized, unweighted, and weighted) of least squares methods, proposed by Ogasawara, for estimating item response theory (IRT) linking coefficients under dichotomous models are extended to the graded response model. A simulation study was conducted to confirm the accuracy of the extended formulas, and a real data study was carried out to compare the performance of the least squares methods with that of moment and characteristic curve linking methods. As found in Ogasawara's study, the generalized least squares method had the smallest asymptotic standard errors but the largest biases of linking coefficient estimates, whereas the unweighted least squares method had the largest asymptotic standard errors but the smallest biases. The weighted least squares method was intermediate. The comparison study showed that with large samples, the weighted least squares method performed nearly as well as the characteristic curve methods in linking accuracy. (Contains 6 tables.) (As Provided).
Anmerkungen	SAGE Publications. 2455 Teller Road, Thousand Oaks, CA 91320. Tel: 800-818-7243; Tel: 805-499-9774; Fax: 800-583-2665; e-mail: journals@sagepub.com; Web site: http://sagepub.com
Erfasst von	ERIC (Education Resources Information Center), Washington, DC
Update	2017/4/10