A Natural B-Spline Varying-Coefficient Method for Longitudinal Binary Response Data with Nonignorable Dropout

• Activity Number: 194
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Jeri Forster*+ and Samantha MaWhinney and Xinshuo M Wang
• Companies: Colorado School of Public Health and Colorado School of Public Health and Colorado School of Public Health
• Keywords: Longitudinal data ; Nonignorable dropout ; Binary ; Varying-coefficient ; HIV

Abstract

Although longitudinal studies provide valuable information, subjects frequently drop out, and the reasons may be related to the outcome of interest. When the probability of dropout depends on unobserved outcomes, even after conditioning on observable data, the missing data are missing not at random and therefore nonignorable. Despite the likelihood of nonignorable dropout, traditional methods are frequently used and likely result in bias and loss of power. Additionally, many times the outcome of interest is binary (e.g., viral load suppression in HIV/AIDS clinical trials). We therefore propose a varying-coefficient mixture model for binary longitudinal data (NSVbin) that allows flexible dependence of the outcome on dropout. The method utilizes natural cubic B-spline basis functions and a lower boundary offset to provide model stability for early dropouts. We present simulations comparing the NSVbin to a binary extension of the conditional linear model where model assumptions are met and violated.