Using R for Estimating Longitudinal Student Achievement Models

Published in: R News, v. 3, no. 3, Dec. 2003, p. 17-23

Posted on on December 01, 2003

by J. R. Lockwood, Harold Doran, Daniel F. McCaffrey

Read More

Access further information on this document at

This article was published outside of RAND. The full text of the article can be found at the link above.

The current environment of test-based accountability in public education has fostered increased interest in analyzing longitudinal data on student achievement. In particular, value-added models (VAM) that use longitudinal student achievement data linked to teachers and schools to make inferences about teacher and school effectiveness have burgeoned. Depending on the available data and desired inferences, the models can range from straightforward hierarchical linear models to more complicated and computationally demanding cross-classified models. The purpose of this article is to demonstrate how R, via the lme function for linear mixed effects models in the nlme package (Pinheiro and Bates, 2000), can be used to estimate all of the most common value-added models used in educational research. After providing background on the substantive problem, the authors develop notation for the data and model structures that are considered. They then present a sequence of increasingly complex models and demonstrate how to estimate the models in R. The authors conclude with a discussion of the strengths and limitations of the R facilities for modeling longitudinal student achievement data.

This report is part of the RAND Corporation External publication series. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.