Abstract:
|
Dropout is a common problem in longitudinal clinical trials and cohort studies, and is of particular concern when dropout occurs for reasons that may be related to the outcome of interest. We present a Bayesian framework for a semi-parametric varying coefficient model for continuous longitudinal data with non-ignorable dropout. The model allows coefficients to vary with dropout time through smooth functions, which are modeled with natural cubic B-splines. The full data is a mixture over the distribution of dropout times, which can be left unspecified. Our Bayesian framework utilizes a reversible jump Markov chain Monte Carlo approach that does not require the choice of a single set of spline knots to make statistical inference, making inference less dependent on "nuisance parameters." These methods reduce bias and more accurately characterize model uncertainty than standard linear mixed models and the conditional linear model to account for dropout. To demonstrate these methods, we present results from a simulation study and estimate the impact of hard drug use on longitudinal CD4+ T cell count in a cohort of untreated HIV infected subjects.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.