JMboost: Boosting joint models

In mayrandy/JMboost: Boosting Joint Models for Longitudinal and Time-to-Event Outcomes

Description

Provides an implementation to boost joint models for longitudinal and time-to-event outcomes.

Usage

JMboost(y, Xl, Xls, last, delta, id, time,
        lambda = 1, alpha = 0.1, mstop_l,
        mstop_ls = NULL, step.length = 0.1,
        betatimeind = 0)

Arguments

y	Longitudinal outcome including observation at or before event (last).
Xl	covariate matrix for longitudinal predictor
Xls	covariate matrix for shared predictor
last	vector indicating if observation refers to longitduinal (FALSE) or shared predictor (TRUE); i.e. if it is the patient's last observation before event.
delta	censoring indicator (person had event or not)
id	1xN vector of subjects
time	all time points (included in Xls if betatimeind != 0)
lambda	starting value baseline hazard
alpha	starting value association parameter
mstop_l	Stopping iteration longitudinal predictor.
mstop_ls	Stopping iteration shared predictor.
step.length	Step length for boosting updates, typical value is 0.1 (default).
betatimeind	indicating which coefficient in shared predictor is time variable (number),

Details

Implements a gradient boosting algorithn to fit joint models considering a longitudinal predictor (w.r.t outcome y) and a shared predictor (affecting the longitudinal outcome and the time-to-event outcome). The relation between the two outcomes is described via the association parameter alpha. For more details, see Waldmann et al. (2016).

References

Waldmann, E., Taylor-Robinson, D., Klein, N., Kneib, T., Pressler, T., Schmid, M., & Mayr, A. (2016). Boosting Joint Models for Longitudinal and Time-to-Event Data. arXiv preprint arXiv:1609.02686.

Examples

set.seed(123)
dat <- simJM(n = 400, n_i = 3, alpha = .5,
            beta = c(1,2,3), betals = c(2,3,1),
            betatimeind = 3, lambda = 0.6)

j1 <- JMboost(y = dat$y, Xl = dat$X, Xls = dat$Xls,
              last = dat$last, delta = dat$delta,
              id = dat$id, time = dat$time, lambda = 1, alpha = 0.1,
              mstop_l = 100, mstop_ls = 100, step.length = 0.1,
              betatimeind = 3)

mayrandy/JMboost documentation built on May 21, 2019, 2:23 p.m.