rpart_exposure: Poisson Recursive Partitioning and Regression Trees with...

In offsetreg: An Extension of 'Tidymodels' Supporting Offset Terms

Poisson Recursive Partitioning and Regression Trees with Exposures

Description

This function is a wrapper around rpart::rpart() for Poisson regression trees using weighted exposures (observation times).

Usage

rpart_exposure(
  formula,
  data,
  exposure_col = "exposure",
  weights = NULL,
  control,
  cost,
  shrink = 1,
  ...
)

Arguments

formula	A model formula that contains a single response variable on the left-hand side.
data	Optional. A data frame containing variables used in the model.
exposure_col	Character string. The name of a column in data containing exposures.
weights	Optional weights to use in the fitting process.
control	A list of hyperparameters. See rpart::rpart.control().
cost	A vector of non-negative costs for each variable in the model.
shrink	Optional parameter for the splitting function. Coefficient of variation of the prior distribution.
...	Alternative input for arguments passed to rpart::rpart.control().

Details

Outside of the tidymodels ecosystem, rpart_exposure() has no advantages over rpart::rpart() since that function allows for exposures to be specified in the formula interface by passing cbind(exposure, y) as a response variable.

Within tidymodels, rpart_exposure() provides an advantage because it will ensure that exposures are included in the data whenever resamples are created.

The formula, data, weights, control, and cost arguments have the same meanings as rpart::rpart(). shrink is passed to rpart::rpart()'s parms argument via a named list. See that function's documentation for full details.

Value

An rpart model

See Also

rpart::rpart()

Examples

if (interactive()) {
  rpart_exposure(deaths ~ age_group + gender, us_deaths,
                 exposure_col = "population")
}

offsetreg documentation built on April 4, 2025, 1:35 a.m.