Hypothesized, Directly-Coded Curve Shapes in Growth Curve Analysis

An Example

Published In: Journal of Methods and Measurement in the Social Sciences, v. 3, no. 2, 2012, p. 13-29

Posted on RAND.org on January 01, 2012

by Patricia M. Herman, Lee Sechrest

Read More

Access further information on this document at journals.uair.arizona.edu

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

Growth curve analysis provides important informational benefits regarding intervention outcomes over time. Rarely, however, should outcome trajectories be assumed to be linear. Instead, both the shape and the slope of the growth curve can be estimated. Non-linear growth curves are usually modeled by including either higher-order time variables or orthogonal polynomial contrast codes. Each has limitations (multicollinearity with the first, a lack of coefficient interpretability with the second, and a loss of degrees of freedom with both) and neither encourages direct testing of alternative hypothesized curve shapes. Especially in studies with relatively small samples it is likely to be useful to preserve as much information as possible at the individual level. This article presents a step-by-step example of the use and testing of hypothesized curve shapes in the estimation of growth curves using hierarchical linear modeling for a small intervention study.

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.