Dr. McNicholas will develop model-based clustering approaches to help discover and understand developmental trajectories of autistic children
McNicholas’ recent work has seen the development of model-based clustering approaches for finding groups of similar observations within multivariate longitudinal data. This work broke new methodological ground but is also important in a variety of real-world applications. One important application of this methodology – and, in fact, the motivating application – is to data from a Canadian longitudinal study on autism in children, where it permits the children to be grouped into developmental trajectories. One advantage of this model-based clustering approach is that the result is a probability that each child belongs to each developmental trajectory. A weakness of this approach is that, while each child has a probability of belonging to each developmental trajectory, children cannot switch between developmental trajectories. To facilitate this additional and practically important flexibility, a new approach will be developed to allow children to switch between trajectories. This novel methodology, which will draw inspiration from a Markov switching model, will lead to new insights into why children switch trajectories. These insights will be highly valuable to clinicians working with autistic children.