Examining the Accuracy of Teacher Value-Added Model Effects Using Bias-Corrected Confidence Intervals

• In Event: Poster Session 10
  In Poster Session: Division D Section 2 Poster Session 2

Abstract

This study demonstrates how to use bias-corrected bootstrapping methods to construct a 95% confidence interval (BCI) for teacher effects in student performance on a high school end-of-course assessment. The teacher effects were estimated based on a context-adjusted two-level hierarchical value-added model (VAM). The sample consisted of 1,434 teachers with 167,488 students who took a QualityCore exam in different subjects between the years of 2010 and 2013. The results show the bias-corrected BCIs have tight ranges around teacher estimates; and about half of teachers with large effects can be statistically distinguished from teachers with lower effects. The methodology demonstrated in this study can be used to provide critical evidence and guidance for policymakers and researchers to evaluate teacher effectiveness.

Authors

• Yincheng Ye, Virginia Polytechnic Institute and State University

• Edgar I. Sanchez, ACT, Inc.