|
Abstract:
|
A value-added model (VAM) is a quasi-experimental statistical model that yields estimates of the contribution of schools, classrooms, teachers, or other educational units to student achievement, controlling for non-school sources of student achievement growth, including prior student achievement, measured student and family characteristics, and (typically) latent student growth trajectories. The objective is to facilitate valid and fair comparisons of productivity with respect to student outcomes, given that schools may serve very different student populations. The conventional value-added model imposes the restriction that a high-performing educational unit is identically high performing for all types of students, including, for example, students with low and high prior achievement and low and high income status. If this assumption is approximately true, classroom and teachers can validly be compared on the basis of a single performance indicator. However, this assumption might be incorrect: A given classroom or teacher could be very effective for students with low prior achievement, for example, but less so with talented and gifted (TAG) students. We present a generalized multivariate value-added model (which we refer to as a differential effects value-added model) that captures differences in value-added productivity (by student subgroups) across schools, classrooms, and teachers (and over time). Multivariate shrinkage is used to produce effect estimates with minimum mean squared error. We illustrate the utility of these models using data from Chicago, Los Angeles, Milwaukee, and New York City.
|