Estimating a meaningful point of change: A comparison of exploratory techniques based on nonparametric regression

Longitudinal studies are increasingly common in psychological research. Characterized by repeated measurements, longitudinal designs aim to observe phenomena that change over time. One important question involves identification of the exact point in time when the observed phenomena begin to meaningfully change above and beyond baseline fluctuations. The authors introduce a nonparametric modeling framework to estimate the change-point of interest using derivatives of the underlying regression function for an outcome variable across time. The estimator of Huh and Carriere (HC) consistently performed acceptably when using a plug-in bandwidth in a Monte Carlo simulation study. It was shown that the estimator performed well with a minimum number of 15 design points. This procedure was applied to longitudinal data collected on performance anxiety. Results suggest that the HC estimator combined with plug-in bandwidth selection provides an efficient strategy for investigating a possible change-point at which an observed outcome variable starts changing across time.