mboost Cox backend

Wrappers for Cox boosting models fitted with mboost.

Usage

fit.mboost.coxph(response, x, cplx, ...)
complexity.cvrisk.mboost(response, x, full.data, max.mstop = 100L, folds = NULL, ...)
# S3 method for class 'mboost_coxph'
predictProb(object, response, x, times, complexity = NULL, ...)
# S3 method for class 'mboost_coxph'
PLL(object, newdata, newtime, newstatus, complexity = NULL, ...)

Arguments

response

survival response as a Surv object or a two-column time/status matrix.

x

covariate matrix.

cplx

selected stopping iteration mstop.

full.data

full data frame, accepted for the peperr complexity-function contract.

max.mstop

maximum number of boosting iterations considered during cross-validation.

folds

optional mboost fold specification passed to cvrisk.

object

a fitted mboost_coxph object.

times

evaluation times for survival probabilities.

complexity

selected stopping iteration.

newdata

new covariate matrix.

newtime

vector of follow-up times.

newstatus

vector of event indicators.

...

additional arguments passed to glmboost, cvrisk, or prediction helpers.

Value

Fitted mboost_coxph objects, selected stopping iterations, survival-probability matrices, and numeric predictive partial log-likelihood values, respectively.

Details

If mboost provides native survival-curve predictions for the fitted model, predictProb.mboost_coxph uses them. Otherwise it falls back to a Breslow baseline estimated from the stored training data and linear predictors.

See also

peperr, glmboost, cvrisk