A tutorial for joint modeling of longitudinal and time-to-event data in R

Cekic, S., Aichele S., Brandmaier, A. M., Köhncke, Y, & Ghisletta, P. Quantitative and Computational Methods in Behavioral Sciences. In press.


In biostatistics and medical research, longitudinal data are often composed of repeated assessments of a variable (e.g., blood pressure or other biomarkers) and dichotomous indicators to mark an event of interest (e.g., recovery from disease, or death). Consequently, joint modeling of longitudinal and time-to-event data has generated much interest in these disciplines over the previous decade. In psychology, too, often we are interested in relating individual trajectories (e.g., cognitive performance or well-being across many years) and discrete events (e.g., death, diagnosis of dementia, or of depression). Yet, joint modeling are rarely applied in psychology and social sciences more generally. This tutorial presents an overview and general framework for joint modeling of longitudinal and time-to-event data, and fully illustrates its application in the context of a behavioral (cognitive aging) study. We discuss practical topics, such as model selection and comparison for both longitudinal and time-to-event data, choice of joint modeling parameterization, and interpretation of model parameters. To do so, we examined seven frequently used packages for joint modeling in the R language and environment. We concluded that of these, JMbayes is especially attractive due to its flexibility, its various parameterizations of the association structure, and for its powerful and fully Bayesian implementation. We make available the R syntax to apply the JMbayes package within our example.

Read the article here.

Published Oct. 1, 2020 3:42 PM - Last modified Mar. 17, 2021 2:12 PM