Date of Completion

5-3-2018

Embargo Period

5-3-2019

Keywords

IRT, Vertical Scaling, Subscore

Major Advisor

H. Jane Rogers

Associate Advisor

Hariharan Swaminathan

Associate Advisor

Christopher Rhoads

Associate Advisor

Tania Huedo-Medina

Associate Advisor

Siang Chee Chuah

Field of Study

Educational Psychology

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

Vertical scaling and subdomain score reporting are two important issues in the current accountability oriented educational environment. They are fundamental for test score reporting to provide evidence on student growth and diagnostic information about special academic needs for students. Even though there is substantial research on both topics, few studies have focused on subdomain score vertical scaling due to the debatable definition of subscales and technical challenges in psychometric models. This dissertation addresses the plausibility of defining subdomain scales from a perspective grounded in cognitive psychology, and employs a two-stage higher-order Item Response Theory (IRT) method for subdomain score vertical scaling in an interpretable and practical manner. Furthermore, this dissertation evaluates the performance of the proposed higher-order IRT method in terms of parameter recovery and investigates the effects on parameter estimation of correlation between higher-order and subdomain traits, subdomain test length, proportion of common items and model identification methods under various simulated conditions. Moreover, this dissertation compares the performance of the proposed higher-order IRT method with the bi-factor IRT model, unidimensional IRT model and score augmentation in vertical scaling. Findings from this dissertation will offer a new perspective for testing and measurement to construct meaningful subdomain scales, and provide a pragmatic and efficient approach for subdomain score vertical scaling.

COinS