Functional Regression in Estimating Post-Traumatic Stress Symptoms Trajectories
Methsarani Premathilaka
Visiting Assistant Professor
Statistical Sciences Department
Wake Forest University
We provide a new approach to analyzing sparse irregularly spaced longitudinal data to study the early development of Post Traumatic Stress Disorder. In our study, we aimed to demonstrate that Functional Data Analysis is a potentially superior tool for analyzing longitudinal trajectories of PTSD symptoms. It has the flexibility to capture the changes in nonlinear patterns and can be used to examine the longitudinal interactions of PTSD symptoms with behavioral, brain structure changes, and other factors. We use data from two existing longitudinal studies and combine together to create a large database. The individual trajectories of PTSD symptoms, depression, and cortical thickness were estimated using Functional Principal component Analysis. Association among these measures was studied using functional regression analysis. Our findings show that the FDA is a powerful tool to study the longitudinal trajectories of the early development of PTSD.