A1280
Title: The growth curve model and multivariate bilinear regression models useful in the analysis of longitudinal data
Authors: Jemila Hamid - University of Ottawa (Canada) [presenting]
Abstract: Inference involving the growth curve model (GCM) will be discussed, which is a generalized multivariate analysis of variance (GMANOVA) model that has been demonstrated to be useful in the analysis of longitudinal data, growth curves, and other response curves. Unlike traditional regression models, the GCM assumes a structured mean (eg. a polynomial mean trajectory over time), which is incorporated through two design matrices: the within and between individual design matrices. Inference involving the mean parameters of the model involve bilinear projections, consequently, the gcm is also referred to as a bilinear regression model. A trace test that was developed as an alternative to the likelihood ratio (LR) test will be presented and its null distribution discussed, including finite sample and asymptotic (large sample) approximations that are useful in practical applications. Results from extensive simulations demonstrating the test's performance will be presented, along with applications to real data sets.
