MachineShop (version 3.8.0)

step_sbf: Variable Selection by Filtering

Description

Creates a specification of a recipe step that will select variables from a candidate set according to a user-specified filtering function.

Usage

step_sbf(
  recipe,
  ...,
  filter,
  multivariate = FALSE,
  options = list(),
  replace = TRUE,
  prefix = "SBF",
  role = "predictor",
  skip = FALSE,
  id = recipes::rand_id("sbf")
)
# S3 method for step_sbf
tidy(x, ...)

Value

An updated version of recipe with the new step added to the sequence of existing steps (if any). For the tidy method, a tibble with columns terms (selectors or variables selected), selected

(logical indicator of selected variables), and name of the selected variable names.

Arguments

recipe

recipe object to which the step will be added.

...

one or more selector functions to choose which variables will be used to compute the components. See selections for more details. These are not currently used by the tidy method.

filter

function whose first argument x is a univariate vector or a multivariate data frame of candidate variables from which to select, second argument y is the response variable as defined in preceding recipe steps, and third argument step is the current step. The function should return a logical value or vector of length equal the number of variables in x indicating whether to select the corresponding variable, or return a list or data frame with element `selected` containing the logical(s) and possibly with other elements of the same length to be included in output from the tidy method.

multivariate

logical indicating that candidate variables be passed to the x argument of the filter function separately as univariate vectors if FALSE, or altogether in one multivariate data frame if TRUE.

options

list of elements to be added to the step object for use in the filter function.

replace

logical indicating whether to replace the original variables.

prefix

if the original variables are not replaced, the selected variables are added to the dataset with the character string prefix added to their names; otherwise, the original variable names are retained.

role

analysis role that added step variables should be assigned. By default, they are designated as model predictors.

skip

logical indicating whether to skip the step when the recipe is baked. While all operations are baked when prep is run, some operations may not be applicable to new data (e.g. processing outcome variables). Care should be taken when using skip = TRUE as it may affect the computations for subsequent operations.

id

unique character string to identify the step.

x

step_sbf object.

See Also

recipe, prep, bake

Examples

library(recipes)

glm_filter <- function(x, y, step) {
  model_fit <- glm(y ~ ., data = data.frame(y, x))
  p_value <- drop1(model_fit, test = "F")[-1, "Pr(>F)"]
  p_value < step$threshold
}

rec <- recipe(rating ~ ., data = attitude)
sbf_rec <- rec %>%
  step_sbf(all_numeric_predictors(),
           filter = glm_filter, options = list(threshold = 0.05))

sbf_prep <- prep(sbf_rec, training = attitude)
sbf_data <- bake(sbf_prep, attitude)

pairs(sbf_data, lower.panel = NULL)

tidy(sbf_rec, number = 1)
tidy(sbf_prep, number = 1)