|
Activity Number:
|
94
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Social Statistics Section
|
| Abstract - #301094 |
|
Title:
|
Estimating Teacher Effects from Longitudinal Data Without Assuming Vertical Scaling
|
|
Author(s):
|
Louis T. Mariano*+ and Daniel McCaffrey and J. R. Lockwood
|
|
Companies:
|
RAND Corporation and RAND Corporation and RAND Corporation
|
|
Address:
|
1200 S. Hayes St, Arlington, VA, 22202,
|
|
Keywords:
|
value added methods ; vertical scaling ; Bayesian hierarchical models ; MCMC ; Wishart prior ; multiple-membership models
|
|
Abstract:
|
There is increasing interest in using longitudinal measures of student achievement to estimate individual teacher effects. Current multivariate models assume scores come from a single vertical scale and that each teacher has a single effect on student outcomes that persists undiminished to all future test administrations (complete persistence) or can diminish with time but remains perfectly correlated (variable persistence). However, vertically linked tests might not be unidimensional, and not all state assessments use a vertical scale. We develop the "generalized persistence" model, a Bayesian multivariate model for estimating teacher effects that accommodates longitudinal data that are not vertically scaled by allowing a teacher's effects on her student's current and future outcomes to have less than perfect correlation. We illustrate the model using mathematics assessment data.
|
