Generic function for extracting th predictive partial log-likelihood from a fitted survival model.
PLL(object, newdata, newtime, newstatus, ...)Vector of length n_new
fitted model of class class.
n_new*p matrix of covariates.
n_new-vector of censored survival times.
n_new-vector of event status, coded with 0 for censoring and 1, if an event occurred.
additional arguments, for example complexity value, if necessary.
The predictive partial log-likelihood measures the prediction performance of each model fitted in a boostrap sample, using the data not in this sample. Multiplying by (-2) leads to a deviance-like measure, which means that small values indicate good prediction performance.
peperr requires function PLL.class in case of survival response, for each model fit of class class.
At the time, PLL.CoxBoost is available.