Incorporating different response formats of competence tests in an IRT model.

Autor/inn/en	Haberkorn, Kerstin; Pohl, Steffi; Carstensen, Claus H.
Titel	Incorporating different response formats of competence tests in an IRT model.
Quelle	In: Psychological test and assessment modeling, 58 (2016) 2, S. 223-252 PDF als Volltext kostenfreie Datei Link als defekt meldenVerfügbarkeit
Beigaben	Literaturangaben
Sprache	englisch
Dokumenttyp	online; gedruckt; Zeitschriftenaufsatz
ISSN	2190-0493; 2190-0507
Schlagwörter	Antwortbogen; Skalierung; Kompetenzmessung; Itemanalyse; Multiple-Choice-Verfahren; Testaufgabe; Dimensionsanalyse
Abstract	Competence tests within large-scale assessments usually contain various task formats to measure the participants' knowledge. Two response formats that are frequently used are simple multiple choice (MC) items and complex multiple choice (CMC) items. Whereas simple MC items comprise a number of response options with one being correct, CMC items consist of several dichotomous true-false subtasks. When incorporating these response formats in a scaling model, they are mostly assumed to be unidimensional. In empirical studies different empirical and theoretical schemes of weighting CMC items in relation to MC items have been applied to construct the overall competence score. However, the dimensionality of the two response formats and the different weighting schemes have only rarely been evaluated. The present study, thus, addressed two questions of particular importance when implementing MC and CMC items in a scaling model: Do the different response formats form a unidimensional construct and, if so, which of the weighting schemes considered for MC and CMC items appropriately models the empirical competence data? Using data of the National Educational Panel Study, we analyzed scientific literacy tests embedding MC and CMC items. We cross-validated the findings on another competence domain and on another large-scale assessment. The analyses revealed that the different response formats form a unidimensional measure across contents and studies. Additionally, the a priori weighting scheme of one point for MC items and half points for each subtask of CMC items best modeled the response formats' impact on the competence score and resembled the empirical competence data well. (Orig.).
Erfasst von	DIPF \| Leibniz-Institut für Bildungsforschung und Bildungsinformation, Frankfurt am Main
Update	2020/2