Different Tests, Different Student Growth Distributions, Different Teacher Effects: Sensitivity of Value-Added Estimation to Test and Scaling

• In Event: Roundtable Session 60
  In Roundtable Session: Approaches to Measuring Growth

Abstract

The purpose of this study is to examine if different tests produce different shapes of student score distribution and how they affect value added estimations. Two different vertically scaled mathematics test scores – a criterion reference and a norm reference test - were used to obtain teacher effect estimates from a large school district. The main findings are as follows. First, different types of tests distinctively describe student growth. Second, the stability of the value added estimation across different tests is low and depends on model selection. Third, the effect of standardization of scores is different in the two tests. These suggest that it is necessary to take test scores’ characteristics or distributions into account when selecting models for VAM.

Authors

• Eun Hye Ham, Michigan State University

• Mark D. Reckase, Michigan State University