workflows (version 1.1.4)

add_variables: Add variables to a workflow

Description

• add_variables() specifies the terms of the model through the usage of tidyselect::select_helpers for the outcomes and predictors.

• remove_variables() removes the variables. Additionally, if the model has already been fit, then the fit is removed.

• update_variables() first removes the variables, then replaces the previous variables with the new ones. Any model that has already been fit based on the original variables will need to be refit.

• workflow_variables() bundles outcomes and predictors into a single variables object, which can be supplied to add_variables().

Usage

add_variables(x, outcomes, predictors, ..., blueprint = NULL, variables = NULL)
remove_variables(x)
update_variables(
  x,
  outcomes,
  predictors,
  ...,
  blueprint = NULL,
  variables = NULL
)
workflow_variables(outcomes, predictors)

Value

• add_variables() returns x with a new variables preprocessor.

• remove_variables() returns x after resetting any model fit and removing the variables preprocessor.

• update_variables() returns x after removing the variables preprocessor, and then re-specifying it with new variables.

• workflow_variables() returns a 'workflow_variables' object containing both the outcomes and predictors.

Arguments

x

A workflow

outcomes, predictors

Tidyselect expressions specifying the terms of the model. outcomes is evaluated first, and then all outcome columns are removed from the data before predictors is evaluated. See tidyselect::select_helpers for the full range of possible ways to specify terms.

...

Not used.

blueprint

A hardhat blueprint used for fine tuning the preprocessing.

If NULL, hardhat::default_xy_blueprint() is used.

Note that preprocessing done here is separate from preprocessing that might be done by the underlying model.

variables

An alternative specification of outcomes and predictors, useful for supplying variables programmatically.

• If NULL, this argument is unused, and outcomes and predictors are used to specify the variables.

• Otherwise, this must be the result of calling workflow_variables() to create a standalone variables object. In this case, outcomes and predictors are completely ignored.

Details

To fit a workflow, exactly one of add_formula(), add_recipe(), or add_variables() must be specified.

Examples

library(parsnip)

spec_lm <- linear_reg()
spec_lm <- set_engine(spec_lm, "lm")

workflow <- workflow()
workflow <- add_model(workflow, spec_lm)

# Add terms with tidyselect expressions.
# Outcomes are specified before predictors.
workflow1 <- add_variables(
  workflow,
  outcomes = mpg,
  predictors = c(cyl, disp)
)

workflow1 <- fit(workflow1, mtcars)
workflow1

# Removing the variables of a fit workflow will also remove the model
remove_variables(workflow1)

# Variables can also be updated
update_variables(workflow1, mpg, starts_with("d"))

# The `outcomes` are removed before the `predictors` expression
# is evaluated. This allows you to easily specify the predictors
# as "everything except the outcomes".
workflow2 <- add_variables(workflow, mpg, everything())
workflow2 <- fit(workflow2, mtcars)
extract_mold(workflow2)$predictors

# Variables can also be added from the result of a call to
# `workflow_variables()`, which creates a standalone variables object
variables <- workflow_variables(mpg, c(cyl, disp))
workflow3 <- add_variables(workflow, variables = variables)
fit(workflow3, mtcars)