The Impact of Functional Form Complexity on Model Overfitting for Intrinsically Nonlinear Longitudinal Data

• In Event: Methodological Considerations and Guidance for Estimating Longitudinal/Growth Models

Abstract

Using nonlinear mixed-effects models (NLMEMs) researchers can model curvilinear patterns of growth. However, there is ambiguity as to what functional form to choose among several nonlinear functions that may fit data equally well. Many of the model fit indices only account for the number of parameters in a model; however, the functional form complexity of these nonlinear models is overlooked. This can lead to models appearing to have good model fit, but are, in reality, capitalizing on variability unique to that specific sample, called “overfitting”. This study aims to assess the performance of model fit indices, whether they adequately capture functional form complexity when assessing model fit, and to suggest less-researched indices that have the potential to capture overfitting.

Authors

• Corissa Taylor Rohloff, University of Minnesota

• Nidhi Kohli

• Seungwon Chung, University of Minnesota