Author : Jeri E. Forster
Publisher :
ISBN 13 : 9781109838824
Total Pages : 178 pages
Book Rating : 4.8/5 (388 download)
Book Synopsis Varying-coefficient Models for Longitudinal Data: Piecewise-continuous, Flexible, Mixed-effects Models and Methods for Analyzing Data with Nonignorable Dropout by : Jeri E. Forster
Download or read book Varying-coefficient Models for Longitudinal Data: Piecewise-continuous, Flexible, Mixed-effects Models and Methods for Analyzing Data with Nonignorable Dropout written by Jeri E. Forster and published by . This book was released on 2006 with total page 178 pages. Available in PDF, EPUB and Kindle. Book excerpt: We address two challenges specific to longitudinal data. Firstly, the relationship between continuous outcome measures and longitudinally collected predictors may vary over time. To flexibly model these relationships and generate interpretable results, we develop mixed-effects, piecewise-continuous varying-coefficient methods (MPV). These techniques will increase goodness-of-fit and hence improve understanding of these dynamic relationships, aiding in hypothesis generation. We apply these MPV models to immunologic and virologic outcome measures collected in HIV/AIDS clinical trials. Secondly, we propose varying-coefficient methods using natural cubic B-spline basis functions (VCM NS) to semiparametrically model the outcome-dropout relationship in clinical trials where nonignorable dropout is present. These methods are computationally stable, highly flexible and relatively simple to implement. Furthermore, we have control over the amount of flexibility applied to the dropout mechanism. As nonignorable dropout frequently exists and naive methods yield biased results, these are valuable qualities for an effective method. We apply the VCM NS and comparable available methods to an HIV/AIDS clinical trial that shows evidence of nonignorable dropout. In addition, we conduct simulation studies to evaluate performance and compare methodologies. The simulation studies suggest that the VCMNS is an improvement over existing methods when the dropout mechanism is nonlinear.