Fitting Curves to Data Series With Seasonality Using the Additive Polynomial Growth Model

• In Event: Poster Session 10
  In Poster Session: Division D Section 2 Poster Session 2

Abstract

Growth curves with seasonal patterns pose a serious challenge to conventional polynomial regression. Familiar drawbacks include serially-correlated errors, higher order terms are hard to interpret, and fit. The situation is often the more acute where the number of measurements are far fewer than a more-sophisticated econometric time-series treatment of seasonality would require. We describe a two-component additive polynomial growth curve model for interim assessments given each fall and spring, in which separate polynomials describe the between-year trend for the fall scores and the between-year changes in fall-to-spring gain scores, jointly. We show that the additive polynomial model provides a better fit to such data and its results are easily interpreted.

Authors

• Yeow Meng Thum, Northwest Evaluation Association

• Tyler H. Matta, University of Oregon