Combines Predictions for Bayesian Model Averaging
Combines estimated survival probabilities or predictions for longitudinal responses.
bma.combine(..., JMlis = NULL, weights = NULL)
...: objects inheriting from class survfit.JMbayes or predict.JMbayes.JMlis: a list of survfit.JMbayes or predict.JMbayes objects.weights: a numeric vector of weights to be applied in each object.an object of class survfit.JMbayes or predict.JMbayes.
Dimitris Rizopoulos d.rizopoulos@erasmusmc.nl
Rizopoulos, D. (2016). The R package JMbayes for fitting joint models for longitudinal and time-to-event data using MCMC. Journal of Statistical Software 72(7) , 1--45. doi:10.18637/jss.v072.i07.
Rizopoulos, D., Hatfield, L., Carlin, B. and Takkenberg, J. (2014). Combining dynamic predictions from joint models for longitudinal and time-to-event data using Bayesian model averaging. Journal of the American Statistical Association
109 , 1385--1397.
## Not run: # we construct the composite event indicator (transplantation or death) pbc2$status2 <- as.numeric(pbc2$status != "alive") pbc2.id$status2 <- as.numeric(pbc2.id$status != "alive") # we fit two joint models using splines for the subject-specific # longitudinal trajectories and a spline-approximated baseline # risk function; the first one with the current value parameterization # and the other with the shared random effects parameterization lmeFit <- lme(log(serBilir) ~ ns(year, 2), data = pbc2, random = ~ ns(year, 2) | id) survFit <- coxph(Surv(years, status2) ~ drug, data = pbc2.id, x = TRUE) jointFit1 <- jointModelBayes(lmeFit, survFit, timeVar = "year") jointFit2 <- update(jointFit1, param = "shared-RE") # we compute survival probabilities for Subject 2 with # different weights ND <- pbc2[pbc2$id == 2, ] # the data of Subject 2 survPreds1 <- survfitJM(jointFit1, newdata = ND, weight = 0.4) survPreds2 <- survfitJM(jointFit2, newdata = ND, weight = 0.6) survPreds.bma <- bma.combine(survPreds1, survPreds2) survPreds.bma plot(survPreds.bma) ## End(Not run)