MachineShop (version 3.9.0)

calibration: Model Calibration

Description

Calculate calibration estimates from observed and predicted responses.

Usage

calibration(
  x,
  y = NULL,
  weights = NULL,
  breaks = 10,
  span = 0.75,
  distr = character(),
  pool = FALSE,
  na.rm = TRUE,
  ...
)

Value

Calibration class object that inherits from data.frame.

Arguments

x

observed responses or resample result containing observed and predicted responses.

y

predicted responses if not contained in x.

weights

numeric vector of non-negative case weights for the observed x responses [default: equal weights].

breaks

value defining the response variable bins within which to calculate observed mean values. May be specified as a number of bins, a vector of breakpoints, or NULL to fit smooth curves with splines for predicted survival probabilities and with loess for others.

span

numeric parameter controlling the degree of loess smoothing.

distr

character string specifying a distribution with which to estimate the observed survival mean. Possible values are "empirical" for the Kaplan-Meier estimator, "exponential", "extreme", "gaussian", "loggaussian", "logistic", "loglogistic", "lognormal", "rayleigh", "t", or "weibull". Defaults to the distribution that was used in predicting mean survival times.

pool

logical indicating whether to compute a single calibration curve on predictions pooled over all resampling iterations or to compute them for each iteration individually and return the mean calibration curve. Pooling can result in large memory allocation errors when fitting smooth curves with breaks = NULL. The current default is changed from versions <= 3.8.0 of the package which only implemented pool = TRUE.

na.rm

logical indicating whether to remove observed or predicted responses that are NA when calculating metrics.

...

arguments passed to other methods.

See Also

c, plot

Examples

# \donttest{
## Requires prior installation of suggested package gbm to run

library(survival)

control <- CVControl() %>% set_predict(times = c(90, 180, 360))
res <- resample(Surv(time, status) ~ ., data = veteran, model = GBMModel,
                control = control)
cal <- calibration(res)
plot(cal)
# }