Date of Award




Document Type


Degree Name

Doctor of Philosophy (PhD)


Department of Educational and Counseling Psychology


Educational Psychology and Methodology

Content Description

1 online resource (xv, 187 pages) : color illustrations.

Dissertation/Thesis Chair

Robert M Pruzek

Committee Members

Heidi Andrade, Kathryn S Schiller


charter schools, multilevel methods, propensity score analysis, propensity scores, school choice, Charter schools, Public schools, Educational accountability, Educational tests and measurements

Subject Categories

Educational Psychology | Statistics and Probability


Unlike their private school counterparts, charter schools receive public funding but are relieved of some of the bureaucratic and regulatory constraints of public schools in exchange for being held accountable for student performance. Studies provide mixed results with regard to charter school performance. Charter schools are, by definition, schools of choice, and this means that observational data methods are required for comparing such schools with others. In observational data contexts, simple comparisons of two groups such as traditional public and charter schools typically ignore the inherent and systematic differences between the two groups. However, given well-designed observational studies and appropriate analysis methods, the effects of the selection bias can be reduced, if not eliminated. The result is that the usual simple comparisons of two independent groups are replaced by comparisons that make adjustments for covariate differences. This study includes development of new methods, largely graphic in form, designed for observational data to compare two groups. These methods are then used to investigate the question of whether students who attend charter schools perform differently than their traditional public school counterparts on two key academic domains: reading and mathematics. The new methods represent extensions of propensity score analysis (Rosenbaum & Rubin, 1983) by aiding descriptions and aim in reducing selection bias in the context of clustered data.