Discipline-Based Education Research Group


Date of this Version


Document Type



Copyright (c) 2014 Walt Stroup


•$1.2 million NSF RETA (Research and Evaluation Technical Assistance), 2011-2014

•Partnership between University of Nebraska-Lincoln (UNL) and Lincoln Public Schools (LPS)

•Focused on developing, evaluating and sharing statistical models to better estimate value-added teacher effects on student learning
“Coherent picture of teaching and learning”

Some Issues Addressed by RETA •Mixed Model Methodology –teacher effects –program effects •Requirements for valid estimates vs real world –models assume •students randomized to teachers •tests do not have ceiling or floor effects –in reality •student assignment not random (for good reasons) •tests often have ceiling / floor effects

Findings •Randomization –previous studies address extreme non-randomization to “game” the VAM –we looked at non-random processes schools actually use –no impact on accuracy, some impact on precision •Ceiling –sufficient impact to invalidate estimates –exacerbated by non-randomization –assessing teacher & program effect requires tests with adequate “stretch”