A Bayesian Bernoulli-Exponential Joint Model for Longitudinal Outcomes and Informative Time Points

• Activity Number: 557 - New Directions in Bayesian Methods for Longitudinal and Graph Data
• Type: Contributed
• Date/Time: Thursday, August 11, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Michael Safo Oduro* and Khalil Shafie
• Companies: University of Northern Colorado and University of Northern Colorado
• Keywords: Longitudinal Data; Bayesian; Informative Time; Model Performance; Bernoulli-Exponential

Abstract

A huge benefit of longitudinal data analysis is its potential to detect changes and observe trends over time. A variety of methods exist for the analysis of this kind of data, majority of which are premised on the assumption of fixed visit time points for study subjects, predetermined by statistical design. This, however, is not always a plausible assumption. Consider, attrition for example, or occurrences that causes changes to the time trajectory of subjects like sickness or adverse events, family emergencies, travel or change of location. These result in not just irregular time points for individuals, but also unbalanced data and differing time/visit profiles for individuals. Visit times, thus can be considered informative, such that subsequent time points for recording observation outcomes of subjects can be adapted based on current subject outcomes. We present a Bayesian model for analyzing joint binary outcomes and varying informative visit times and subsequently investigate the influence of controlled variations in visit, prior and sample size schemes on model performance.