Attitudinal Data: Dimensionality and Start Values for Estimating Item Parameters.

Autor/inn/en	Nandakumar, Ratna; Hotchkiss, Larry; Roberts, James S.
Titel	Attitudinal Data: Dimensionality and Start Values for Estimating Item Parameters.
Quelle	(2002), (29 Seiten) PDF als Volltext kostenfreie Datei Verfügbarkeit
Sprache	englisch
Dokumenttyp	gedruckt; online; Monographie
Schlagwörter	Attitudes; Estimation (Mathematics); Simulation; Test Items + Suchen Sie Ihr Suchwort? Attitude; Einstellung; Verhalten; Estimation; Mathematics; Schätzung; Simulation program; Simulationsprogramm; Test content; Testaufgabe
Abstract	The purpose of this study was to assess the dimensionality of attitudinal data arising from unfolding models for discrete data and to compute rough estimates of item and individual parameters for use as starting values in other estimation parameters. One- and two-dimensional simulated test data were analyzed in this study. Results of limited analyses performed so far have shown that linear principal components analysis of unfolding data provides a reliable estimate of the underlying dimensionality. For every unfolding dimension, there are two linear principal components. In addition, pattern coefficients of items on the two principal components associated with each dimension form a fan-shaped simplex pattern resembling a semicircle. Arc length of an item along the semicircle can serve as an estimate of item location. Length of an item in the item space spanned by the two linear components can serve as an estimate of item discrimination. For individuals, arc lengths computed using the individual scores on the two principal components could serve as estimates of individuals' location parameters. For two dimensional test data, an algorithm has been developed to identify component pairs associated with each dimension and to classify items correctly into dimensional groups. (Contains 1 table, 7 figures, and 16 references.) (Author/SLD)
Erfasst von	ERIC (Education Resources Information Center), Washington, DC