sambrilleman/rstanjm: Bayesian Joint Longitudinal and Time-to-Event Models via Stan

Estimate joint longitudinal and time-to-event models under a Bayesian framework via Stan. Stan is a C++ library for Bayesian estimation. Users of the rstanjm package specify a joint model using the customary R formula syntax and customary R data frames. The rstanjm package then carries out back-end estimation in Stan via the 'rstan' package (the R interface for Stan). Univariate (one longitudinal marker) or multivariate (more than one longitudinal marker) joint models are allowed. The longitudinal outcome(s) are each modelled using a generalised linear mixed model. Dependence between multiple longitudinal outcomes is captured via a shared multivariate normal distribution for the random effects. Continuous, binary or count data can be handled in the longitudinal submodel via a range of link functions and error distributions. Multilevel clustered data (for example, patients within clinics) can be accomodated in the longitudinal submodel only. The time-to-event is modelled using a proportional hazards model for which the baseline hazard can be specified as a Weibull distribution, piecewise constant, or approximated using B-splines. The association structure for the joint model can be specified in a variety of ways, for example, by having the log hazard of the event linearly associated with the current value of the longitudinal marker, current slope of the longitudinal marker, or shared random effects. Various options are available for the prior distributions of the regression coefficients.

Getting started

Package details

LicenseGPL (>=3)
Package repositoryView on GitHub
Installation Install the latest version of this package by entering the following in R:
sambrilleman/rstanjm documentation built on May 25, 2017, 11:39 p.m.