Joint models for a GLM-type longitudinal response and a time-to-event with smooth random effects
Longitudinal studies often entail non-Gaussian primary responses. When dropout occurs, potential non-ignorability of the missingness process may occur, and a joint model for the primary response and a time-to-event may represent an appealing tool to account for dependence between the
two processes. As an extension to the GLMJM, recently proposed, and based on Gaussian latent effects, we assume that the random effects follow a smooth, P-spline based density. To estimate model parameters, we adopt a two-step conditional Newton–Raphson algorithm. Since the maximization
of the penalized log-likelihood requires numerical integration over the random effect, which is often cumbersome, we opt for a pseudo-adaptive Gaussian quadrature rule to approximate the model likelihood. We discuss the proposed model by analyzing an original dataset on dilated cardiomyopathies
and through a simulation study.
Keywords: 01A23; 45B67; Joint models; MDCM data; mixture of P-splines; smooth random effect distribution
Document Type: Research Article
Affiliations: Statistics Division, Food and Agriculture Organization (FAO) of the United Nations, Viale delle Terme di Caracalla, 00153, Roma, Italy
Publication date: 18 November 2019
- Access Key
- Free content
- Partial Free content
- New content
- Open access content
- Partial Open access content
- Subscribed content
- Partial Subscribed content
- Free trial content