Advances and Applications in Statistics
Volume 64, Issue 1, Pages 33 - 62
(September 2020) http://dx.doi.org/10.17654/AS064010033 |
|
FITTING JOINT MODELING OF LONGITUDINAL AND TIME-TO-EVENT DATA USING STOCHASTIC EM APPROACH
Dina M. Sabry, Ahmed M. Gad and Ramadan H. Mohamed
|
Abstract: Most medical studies produce two different types of data for each study subject: longitudinal and time-to-event data. Joint analysis is an appealing approach that can model the association between an event of interest and a time dependent covariate that is measured with error. Although primarily developed for this purpose, joint models can also be used as an efficient tool to handle non-ignorable missingness in longitudinal studies. In this paper, we adopt a joint model to handle both monotone and non-monotone missingness simultaneously. The stochastic EM (SEM) algorithm is proposed and developed to obtain the parameter estimates. The Monte Carlo method is developed to compute the standard errors of the estimates. The proposed approach is evaluated and assessed using a simulation study. Also, the approach is applied to a real data from a clinical trial for the Scleroderma lung disease. |
Keywords and phrases: stochastic EM algorithm, shared parameter model, joint modeling, longitudinal data, survival data, linear mixed effects model, Cox model.
|
|
Number of Downloads: 205 | Number of Views: 643 |
|