An Investigation of the Behavior of Value-Added Models for Estimating MSP Impact
Abstract
"Appropriate evaluation of MSP programs is essential. One tool for measuring the impact of MSP programs on teaching effectiveness is student achievement data. This requires a coherent picture of student progress before, during, and after a project's professional development program. However, the data available to projects often do not meet the technical requirements of current statistical methods including value-added models (VAMs). Specifically, models assume random assignment of students to classes when, in reality, students are not randomly assigned to classes for a variety of educationally legitimate reasons. This project is investigating the impact of non-randomization on the ability of VAM to detect impact of an MSP project. The examination of non-randomization will be presented and discussed in this session."
