Title: Overcoming challenges in the analysis of longitudinal discrete data
Authors: Justine Shults - Perelman School of Medicine at University of Pennsylvania and Children's Hospital of Philadelphia (United States) [presenting]
Abstract: Challenges are faced in the analysis of longitudinal discrete data that we do not face with continuous outcomes. Likelihoods for discrete outcomes can be complex, especially for longitudinal data with serial correlation and overdispersion. Semi-parametric approaches such as generalized estimating equations (GEE) are appealing because they do not require specification of the full likelihood. However, it is possible to unknowingly obtain estimates for which there is no valid parent distribution. We demonstrate some of the challenges we face in the analysis of longitudinal discrete data and show how we can overcome some of the difficulties via the first-order Markov maximum likelihood based approach.