Files in this item



application/pdf9717348.pdf (3MB)Restricted to U of Illinois
(no description provided)PDF


Title:Some issues in item response theory
Author(s):Wu, Hongsheng
Doctoral Committee Chair(s):Stout, William F.
Department / Program:Statistics
Degree Granting Institution:University of Illinois at Urbana-Champaign
Abstract:The issues studied in this thesis are: the assumption of local independence (LI) in item response theory (IRT) models in the first three chapters, and differential item functioning (DIF) in IRT in the last chapter. A new term "conditional trivariance" for three items is defined, which is a measurement of joint local dependence among item triples. This is extended to any J items. Relevant theorems are stated and proved, which provide necessary and sufficient conditions involving conditional co-variances, trivariances, etc. for local independence to hold. These theorems lead us to a method to test LI that goes beyond conditional covariance exploration. We realize this procedure by using a kernel smoothing technique. Asymptotic normality of the test statistic is given and proved, which in turn justifies our procedure. Simulation studies show the procedure works well for the case of large examinee sample sizes. Some Graduate Record Examination (GRE) verbal test data is analyzed.
Chapter 4 deals DIF. A two-stage kernel smoothed DIF detecting method is proposed. Two-stage method is obtained by the amalgamation both Douglas, J., Stout, W., & DiBello, L.'s kernel smoothed DIF detection method and Shealy & Stout's linear regression correction used in SIBTEST. It is compared with Douglas, et al.'s one-stage method. The factors influencing the performance of the procedures are studied. A best combination of these factors is recommended.
Issue Date:1996
Rights Information:Copyright 1996 Wu, Hongsheng
Date Available in IDEALS:2011-05-07
Identifier in Online Catalog:AAI9717348
OCLC Identifier:(UMI)AAI9717348

This item appears in the following Collection(s)

Item Statistics