Longitudinal K-sets analysis using lagged variables

Abstract

We present an application of nonlinear Generalised Canonical Analysis (GCA) for analysing longitudinal data. The application uses lagged versions of variables to accomodate the time-dependence in the measurements. The usefulness of the proposed method is illustrated in an example from developmental psychology, in which we explore the relationship between mother and child dyadic interaction during the first six months after birth, demonstrating how child behaviour can elicit mother behaviour. We discuss the relationship between our proposed method and the most closely resembling SERIALS (Van Buuren, 1990) method for nonlinear time series analysis.