Please find below details of the next meeting, everybody welcome.
Royal Statistical Society Merseyside Local Group
Wednesday 27th June 2007 2.00pm
(Tea & Coffee in Common room afterwards, 3rd Floor)
Venue : Penthouse, Maths & Oceanography building University of Liverpool
Please contact Ashley Jones ([log in to unmask]) for further details.
---------------------------
Joint modelling of Longitudinal and Survival outcomes
Ruwanthi Kolamunnage-Dona (University of Liverpool)
Ines Sousa (Lancaster University)
Peter Philipson (Newcastle University)
This work formulates a class of models for the joint behaviour of longitudinal measurements and associated time to event outcomes. Standard joint modelling of longitudinal and survival data have one failure type and an assumption of independent censoring. The implementation of such models in practice is discussed, with particular attention paid to the development of easy-to-use R software for the case of a solitary survival outcome.
We extend the standard joint model to analyse informative competing risks. A cause-specific hazards model is fitted to allow for competing risks, with a separate association parameter for each competing risk, and a combined analysis is undertaken. This model is applied to simultaneously investigate the differential treatment effects on competing risks of drug withdrawal due to intolerable side effects or poor seizure control and longitudinal measurements of quality of life/anti-epileptic drug dose in newly-diagnosed patients with epilepsy.
The model is extended further to a joint analysis of multivariate repeated measurements and multiple time-to-event outcomes. We propose a multivariate extension of the fully parametric transformed Gaussian joint model. The model assumes a multivariate Gaussian joint distribution for the repeated measurement variables and log-transformed event times. Inference is straightforward and the likelihood function remains tractable when an event acts either to terminate the sequences (informative dropout) or to censor other time-to-event outcomes (competing risks).
|