This repository contains the source files for the R package JMbayes. This package fits joint models for longitudinal and time-to-event data under a Bayesian approach using MCMC. These models are applicable in mainly two settings. First, when focus is on the survival outcome and we wish to account for the effect of an endogenous (aka internal) time-dependent covariates measured with error. Second, when focus is on the longitudinal outcome and we wish to correct for nonrandom dropout.
The package contains two main joint-model-fitting functions, jointModelBayes() and mvJointModelBayes() with similar syntax but different capabilities.
jointModelBayes()It can fit joint models for a single longitudinal outcome and a time-to-event outcome.
The user can specify her own density function for the longitudinal responses using argument densLong (default is the normal pdf). Among others, this allows to fit joint models with categorical and left-censored longitudinal responses and robust joint models with Student’s-t error terms. In addition, using the df.RE argument, the user can also change the distribution of the random effects from multivariate normal to a multivariate Student’s-t with prespecified degrees of freedom.
For the survival outcome a relative risk models is assumed with a B-spline approximation for the baseline hazard (penalized (default) or regression splines can be used). Left-truncation and exogenous time-varying covariates can also be accommodated.
The user has now the option to define custom transformation functions for the terms of the longitudinal submodel that enter into the linear predictor of the survival submodel (arguments extraForm, param). For example, the current value of the longitudinal outcomes, the velocity of the longitudinal outcome (slope), the area under the longitudinal profile. From the aforementioned options, in each model up to two terms can be included. In addition, using argument transFun interactions terms, nonlinear terms (polynomials, splines) can be considered.
mvJointModelBayes()It can fit joint models for multiple longitudinal outcomes and a time-to-event outcome.
The longitudinal part of the joint model is a multivariate generalized linear mixed effects models, currently allowing for normal, binary and Poisson outcomes. This model is first fitted using function mvglmer().
For the survival outcome a relative risk models is assumed with a B-spline approximation for the baseline hazard (penalized (default) or regression splines can be used). Left-truncation, interval censored data and exogenous time-varying covariates can also be accommodated.
The user has now the option to define custom transformation functions for the terms of the longitudinal submodel that enter into the linear predictor of the survival submodel (argument Formulas). For example, the current value of the longitudinal outcomes, the velocity of the longitudinal outcome (slope), the area under the longitudinal profile. From the aforementioned options, in each model limitless terms can be included. In addition, using argument Interactions allows to include interactions terms of the longitudinal components with other observed factors. A special case for this argument is to use function tve() that allows for time-varying regression coefficients in the relative risk model. Furthermore, argument transFuns allows to transform the longitudinal components using some pre-defined transformation function (i.e., exp(), expit(), log, sqrt()).
The aforementioned features are illustrated in the Multivariate Joint Models vignette.
Function survfitJM() computes dynamic survival probabilities.
Function predict() computes dynamic predictions for the longitudinal outcome.
Function aucJM() calculates time-dependent AUCs for joint models, and function rocJM() calculates the corresponding time-dependent sensitivities and specifies.
Function prederrJM() calculates prediction errors for joint models.
Function runDynPred() invokes a shiny application that can be used to streamline the calculation of dynamic predictions for models fitted by JMbayes.
Vignettes are available in the doc directory:
Multivariate_Joint_Models.html illustrates the basic capabilities of mvJointModelBayes().
Dynamic_Predictions.html illustrates how dynamic predictions from multivariate joint models can be computed and evaluated.