Bootstrap Error Rate Estimators

Those functions are low-level functions used by errorest and are normally not called by users.

Usage

bootest(y, ...)
# S3 method for class 'factor'
bootest(y, formula, data, model, predict, nboot=25,
bc632plus=FALSE, list.tindx = NULL, predictions = FALSE, 
both.boot = FALSE, ...)

Arguments

y

the response variable, either of class factor (classification), numeric (regression) or Surv (survival).

formula

a formula object.

data

data frame of predictors and response described in formula.

model

a function implementing the predictive model to be evaluated. The function model can either return an object representing a fitted model or a function with argument newdata which returns predicted values. In this case, the predict argument to errorest is ignored.

predict

a function with arguments object and newdata only which predicts the status of the observations in newdata based on the fitted model in object.

nboot

number of bootstrap replications to be used.

bc632plus

logical. Should the bias corrected version of misclassification error be computed?

predictions

logical, return a matrix of predictions. The ith column contains predictions of the ith out-of-bootstrap sample and 'NA's corresponding to the ith bootstrap sample.

list.tindx

list of numeric vectors, indicating which observations are included in each bootstrap sample.

both.boot

logical, return both (bootstrap and 632plus) estimations or only one of them.

...

additional arguments to model.

Details

See errorest.