Linear Mixed Joint Model with Trigonometric Basis for Longitudinal Outcomes

• Activity Number: 418
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Scientific and Public Affairs Advisory Committee
• Author(s): Peng Wu* and Yuanjia Wang and Bernard Timothy Walsh and Evelyn Attia
• Companies: Columbia University and Columbia University and Columbia University Medical Center and Columbia University Medical Center
• Keywords: Linear mixed model ; Trigonometric basis ; Joint modeling ; Physical activity

Abstract

We propose a linear mixed joint model approach using trigonometric basis to study two longitudinal responses in our work. Our method is motivated by a functional physical activity study from New York State Psychiatric Institute in which 63 female patients with Anorexia Nervosa were enrolled and prospectively followed for 3 study sessions in order to study physical activity. Repeated measures of activity count were recorded by the Intelligent Device for Energy Expenditure and Activity (IDEEA) device and follow-up body mass index (BMI) in the 12 months after discharge from inpatient treatment was computed. It is of great scientific interest to investigate the covariate effects on both outcomes. To this end, we propose a joint modeling method where linear mixed models with trigonometric basis are utilized to account for 24-hour cyclic physical activity patterns and to estimate covariate effects on both outcomes as well as on within-subject association. Our method is applied to IDEEA data.