New nonparametric statistical procedures for analyzing bias/DIF and dimensionality in item response data

Unidimensionality is one of the most important assumptions required by much of the currently used item response theory (IRT) methodologies. In the first part of this thesis, a further and non-trivial practical refinement of DIMTEST(Stout, 1987; Nandakumar & Stout, 1993) is made to assess latent trait unidimensionality for mixed dichotomous and polytomous items. The modification is referred to Poly-DIMTEST. The new test statistic for polytomous item scoring was carefully developed and defended with an appropriate asymptotic theory. A simulation study then was carried out to investigate the performance of Poly-DIMTEST. The results demonstrate that Poly-DIMTEST has good Type I error as well as good power. We conclude that the Poly-DIMTEST procedure shows promise as a useful tool in assessing unidimensionality for mixed dichotomous and polytomous test data.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Statistics

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2011

Author dc:creator

• Li, Hsin-Hung

Contributors dc:contributor

• Stout, William F.

• Education, Tests and Measurements
• Statistics
• Psychology, Psychometrics

Statement dc:rights

• Copyright 1995 Li, Hsin-Hung

Language dc:language

eng

Identifier

AAI9543649
(UMI)AAI9543649

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/22859