Date of Award

Summer 8-2015

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Educational Studies and Research

Committee Chair

Dr. Kyna Shelley

Committee Chair Department

Educational Studies and Research

Committee Member 2

Dr. Thomas J. Lipscomb

Committee Member 2 Department

Educational Studies and Research

Committee Member 3

Dr. Lilian H. Hill

Committee Member 3 Department

Educational Studies and Research

Committee Member 4

Dr. Richard S. Mohn

Committee Member 4 Department

Educational Studies and Research

Abstract

The use of test-based accountability has expanded beyond measurements of school effectiveness to include measurements of teacher effectiveness. However, whereas the use of test-based accountability has expanded, the understanding of the statistical methodologies used in accountability systems has not kept pace. Currently, Student Growth Percentiles and value-added modeling are the most prevalent methodologies for estimating annual student growth. Each of these methodologies is regression-based and relies on scale scores from standardized assessments. Given the prevalence of Item Response Theory in statewide assessment programs, these scale scores often result from Item Response Theory scaling practices. Grounded in earlier work of Brockman (2011), Chiu and Camilli (2013), and Lu, Thomas, and Zumbo (2005), concerning error related to Item Response Theory-based scale scores, this study considers using Item Response Theory as the measurement model in a structural equation model by including simulated item response patterns as indicators of ability. Data were simulated using parameters from the Mississippi Curriculum Test, Second Edition. Separate structural equation models for language arts and mathematics were considered. Upon examining the fit of each model, results indicated a good fit for the measurement model in language arts and in mathematics. Results also indicated a good fit for the overall structural equation model but none of the structural relationships were statistically significant. Additional results and implications of this study are discussed.

Share

COinS