Tuesday 28th April 2015
Seminar: Life-Course Data Analysis: A Generalised Hierarchical Unimodal Curve Registration Approach
Speaker: Elena Erosheva, Centre for Statistics and Social Sciences, University of Washington
Time: 13.00 start
Location: COL 615 (Leverhulme Library, Columbia House)
Abstract: A major aim of life course data analysis is to describe the within- and between-individual variability in a behavioural outcome, such as crime or drug use. Popular methods for analysing individual behavioural trajectories -- group-based trajectory and growth mixture models -- rely on a combination of growth curve and mixture modeling. Applications of these methods have been popular with criminologists and developmental psychologists, and are quickly expanding to other areas such as clinical medicine and public health. Most commonly, these applications seek to identify distinct trajectory patterns. In this talk, I start by describing these traditional approaches to life course data analysis, and summarize observations on the statistical practice. I then present an alternative approach, Unimodal Curve Registration (UCR). This approach allows researchers to describe individual criminal careers via individual-specific phase and amplitude departures from the unimodal population age-crime curve. The generalized hierarchical UCR approach accommodates different types of outcomes via underlying latent Gaussian random variables, and allows for covariate effects on phase and amplitude. The ideas are illustrated using (a) scale data on antisocial behaviour from Montreal Longitudinal and Experimental Study and (b) self-reported counts of yearly marijuana use from the Denver Youth Survey.