Evaluate the Brier score, i.e. prediction error, for a fitted model on new data. To be used as argument aggregation.fun in peperr call.
aggregation.brier(full.data=NULL, response, x, model, cplx=NULL,
type=c("apparent", "noinf"), fullsample.attr = NULL, ...)Scalar, indicating the empirical Brier score.
passed from peperr, but not used for calculation of the Brier score.
vector of binary response.
n*p matrix of covariates.
model fitted as returned by a fit.fun, as used in a call to peperr.
passed from peperr, but not necessary for calculation of the Brier score.
character.
passed from peperr, but not necessary for calculation of the Brier score.
additional arguments, passed to predict function.
The empirical Brier score is the mean of the squared difference of the risk prediction and the true value of all observations and takes values between 0 and 1, where small values indicate good prediction performance of the risk prediction model.