Continuous Trajectory Modeling When Data Are Collected Longitudinally at Common Discrete Time Points

• Activity Number: 162
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): John Boscardin*+
• Companies: San Francisco VA/University of California at San Francisco
• Keywords: mixed effects models ; spline models ; non-normal data distributions

Abstract

Detailed modeling of continuous trajectories for longitudinal data is often hampered by a small number of discrete time points for data collection. Changing the time scale of the analysis can present a solution. We discuss two classes of examples: (i) using age of the subject instead of time on study, and (ii) centering the followup time around an event time for which exact date of occurrence is available (e.g. date of hospitalization). The examples will be discussed in the context of a longitudinal study on older adults where most measures are collected at regular bi-annual intervals but exact dates are available for a number of events of interest. Our interest lies in modeling pre- and post-event trajectories of a longitudinal measure. Informative dropout due to subject death is accounted for using a joint model of the longitudinal and mortality data.