add_case_weights: Add case weights to a workflow

View source: R/pre-action-case-weights.R

add_case_weightsR Documentation

Add case weights to a workflow

Description

This family of functions revolves around selecting a column of data to use for case weights. This column must be one of the allowed case weight types, such as hardhat::frequency_weights() or hardhat::importance_weights(). Specifically, it must return TRUE from hardhat::is_case_weights(). The underlying model will decide whether or not the type of case weights you have supplied are applicable or not.

  • add_case_weights() specifies the column that will be interpreted as case weights in the model. This column must be present in the data supplied to fit().

  • remove_case_weights() removes the case weights. Additionally, if the model has already been fit, then the fit is removed.

  • update_case_weights() first removes the case weights, then replaces them with the new ones.

Usage

add_case_weights(x, col)

remove_case_weights(x)

update_case_weights(x, col)

Arguments

x

A workflow

col

A single unquoted column name specifying the case weights for the model. This must be a classed case weights column, as determined by hardhat::is_case_weights().

Details

For formula and variable preprocessors, the case weights col is removed from the data before the preprocessor is evaluated. This allows you to use formulas like y ~ . or tidyselection like everything() without fear of accidentally selecting the case weights column.

For recipe preprocessors, the case weights col is not removed and is passed along to the recipe. Typically, your recipe will include steps that can utilize case weights.

Examples

library(parsnip)
library(magrittr)
library(hardhat)

mtcars2 <- mtcars
mtcars2$gear <- frequency_weights(mtcars2$gear)

spec <- linear_reg() %>%
  set_engine("lm")

wf <- workflow() %>%
  add_case_weights(gear) %>%
  add_formula(mpg ~ .) %>%
  add_model(spec)

wf <- fit(wf, mtcars2)

# Notice that the case weights (gear) aren't included in the predictors
extract_mold(wf)$predictors

# Strip them out of the workflow, which also resets the model
remove_case_weights(wf)

workflows documentation built on May 29, 2024, 3:57 a.m.