Presenter: Luis Ledesma

Supervisory Committee: Eleanor Pullenayegum (Supervisor), Aya Mitani and Kuan Liu
Date and Time: Monday, September 11th, 2023 at 2pm EST

Location: Health Sciences Building (155 College St), Room HS650

Zoom: https://utoronto.zoom.us/j/87849654795

Abstract: Longitudinal data is often subject to irregular visiting times. While there are established methods developed to deal with certain irregular visiting patterns, not much is known regarding modelling scenarios where the current value of the outcome influences the visit intensity at that time. One model that allows for such dependence provided there is conditional independence given a random effect is the Sun (2011) multiplicative model, which is one of the few semi-parametric joint models that is suitable for count data. We extend this model by additionally estimating the intercept term which was formerly treated as a nuisance parameter. This technique enables us to assess prognosis by examining the evolution of the outcome variable over time, rather than just the relative effects of model covariates. This extended estimator exhibits desirable large-sample properties such as consistency and asymptotic normality. We evaluate finite-sample performance with a simulation study, which shows that the proposed method outperforms standard approaches in terms of bias and standard error under a range of scenarios. Moreover, the proposed model is more computationally efficient than the standard method for different specifications of the intercept term. Furthermore, we present an application to a study of tumour recurrence for a longitudinal cohort of bladder cancer patients characterized by informative visits. We recommend that unless the intercept term fluctuates extensively through time, applying our proposed method results in improved model estimates and reduced computational complexity compared to the standard Sun (2011) model estimation, and permits the additional estimation of the mean outcome trajectories.

All are welcome!