A Joint Model for Longitudinal Responses with Missing Data

• Activity Number: 193
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Brenden Bishop*
• Companies: The Ohio State University
• Keywords: Multivariate Analysis ; Longitudinal ; Joint Model ; Missing Data ; Psychometrics

Abstract

Assessing the long-term impact of a clinical intervention requires longitudinal data. When assessing individuals over time, encountering missing data is likely. Joint models are commonly used in biostatistics and epidemiology to jointly model both responses and missingness, particularly when missingness is due to participant mortality. By contrast, in psychology joint models infrequently appear. This work presents an application of a joint model for longitudinal responses and missingness to a dataset in which the proportion of missingness increases curvilinearly over time. The data consist of psychological measures of mood, cancer related stress, and depression as well as disease relevant covariates from a sample of U.S. cancer patients undergoing a novel immunotherapy or chronic lymphocytic leukemia. A research team comprised of clinical psychologists and medical oncologists were interested in the change over the course of treatment of psychological variables. A joint model is illustrated to account for any bias in the parameter estimates due to missing data.