bsPDP.caret: Bootstrapped partial dependence plots for 'caret' functions
In bangecon/bsPDP: Bootstrapped PDPs with Confidence Intervals

Description Usage Arguments Value References

Evaluates a models in the caret package for n_boot bootstrap iterations to obtain confidence bands for the estimated PDP.

bsPDP.caret(
  formula,
  variable = NULL,
  data,
  newdata = NULL,
  grid = NULL,
  outcome = NULL,
  n_boot = 100,
  p_boot = 0.6,
  N = 1000,
  label = NULL,
  clock = FALSE,
  test = FALSE,
  seed = 8675309,
  trainMethod = "rf",
  ...
)

`formula`	is an output formula for the outcome.
`variable`	is the treatment variable.
`data`	is a data frame to be used for the training.
`newdata`	is an optional data frame of test data for the PDPs.
`grid`	sets the values of `variable` to evaluate the default is 100 values in range of `variable`: grid = seq(min(x), max(x), length.out = 100).
`n_boot`	is the number of bootstrap replications.
`p_boot`	is the proportion of the data to select for each bootstrap replication.
`N`	is the number of observations to select for calculating the PDPs.
`label`	is a character-string variable label for `variable`.
`clock`	is a logical indicating whether to time each bootstrap replication.
`test`	is a logical indicating whether to calculate the pdp for both the `data` and `newdata`.
`seed`	is a random seed (default is 8675309).
`trainMethod`	is the train method for the estimation default is `"rf"` (random forest). For details, see http://topepo.github.io/caret/using-your-own-model-in-train.html.
`...`	additional arguments for the `train` function, see `?train` or http://topepo.github.io/caret/using-your-own-model-in-train.html. We recommended checking settings for `trControl = trainControl()` (control the computational nuances of the train function, see ?trainControl) and `tuneGrid` (a data frame with possible tuning values).

bsPDPcaret returns an object with class "bsPDP," a list that includes the following components:

`variable`	the treatment variable.
`pdpData`	the estimated average predictions and standard errors along `variable`.
`trainData`	the original training data.
`testData`	the test data.
`outcome`	the outcome class (`NULL` for outcomes with `numeric` class).
`method`	the method (`c("caret", method)`)