step_profile: Create a Profiling Version of a Data Set

Description

step_profile creates a specification of a recipe step that will fix the levels of all variables but one and will create a sequence of values for the remaining variable. This step can be helpful when creating partial regression plots for additive models.

Usage

step_profile(recipe, ..., profile = NULL, pct = 0.5, index = 1,
  grid = list(pctl = TRUE, len = 100), columns = NULL, role = NA,
  trained = FALSE, skip = FALSE, id = rand_id("profile"))
# S3 method for step_profile
tidy(x, ...)

Arguments

recipe

A recipe object. The step will be added to the sequence of operations for this recipe.

...

One or more selector functions to choose which variables will fixed to a single value. See selections() for more details. For the tidy method, these are not currently used.

profile

A call to dplyr::vars()) to specify which variable will be profiled (see selections()). If a column is included in both lists to be fixed and to be profiled, an error is thrown.

pct

A value between 0 and 1 that is the percentile to fix continuous variables. This is applied to all continuous variables captured by the selectors. For date variables, either the minimum, median, or maximum used based on their distance to pct.

index

The level that qualitative variables will be fixed. If the variables are character (not factors), this will be the index of the sorted unique values. This is applied to all qualitative variables captured by the selectors.

grid

A named list with elements pctl (a logical) and len (an integer). If pctl = TRUE, then len denotes how many percentiles to use to create the profiling grid. This creates a grid between 0 and 1 and the profile is determined by the percentiles of the data. For example, if pctl = TRUE and len = 3, the profile would contain the minimum, median, and maximum values. If pctl = FALSE, it defines how many grid points between the minimum and maximum values should be created. This parameter is ignored for qualitative variables (since all of their possible levels are profiled). In the case of date variables, pctl = FALSE will always be used since there is no quantile method for dates.

columns

A character string that contains the names of columns that should be fixed and their values. These values are not determined until prep.recipe() is called.

role

Not used by this step since no new variables are created.

trained

A logical to indicate if the quantities for preprocessing have been estimated.

skip

A logical. Should the step be skipped when the recipe is baked by bake.recipe()? While all operations are baked when prep.recipe() is run, some operations may not be able to be conducted on new data (e.g. processing the outcome variable(s)). Care should be taken when using skip = TRUE as it may affect the computations for subsequent operations

A character string that is unique to this step to identify it.

A step_profile object.

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 (which is the columns that will be affected), and type (fixed or profiled).

Details

This step is atypical in that, when baked, the new_data argument is ignored; the resulting data set is based on the fixed and profiled variable's information.

Examples

Run this code

# NOT RUN {
data(okc)

# Setup a grid across date but keep the other values fixed
recipe(~ diet + height + date, data = okc) %>%
  step_profile(-date, profile = vars(date)) %>%
  prep(training = okc, retain = TRUE) %>%
  juice


##########

# An *additive* model; not for use when there are interactions or
# other functional relationships between predictors

lin_mod <- lm(mpg ~ poly(disp, 2) + cyl + hp, data = mtcars)

# Show the difference in the two grid creation methods

disp_pctl <- recipe(~ disp + cyl + hp, data = mtcars) %>%
  step_profile(-disp, profile = vars(disp)) %>%
  prep(training = mtcars, retain = TRUE)

disp_grid <- recipe(~ disp + cyl + hp, data = mtcars) %>%
  step_profile(
    -disp,
    profile = vars(disp),
    grid = list(pctl = FALSE, len = 100)
  ) %>%
  prep(training = mtcars, retain = TRUE)

grid_data <- juice(disp_grid)
grid_data <- grid_data %>%
  mutate(pred = predict(lin_mod, grid_data),
         method = "grid")

pctl_data <- juice(disp_pctl)
pctl_data <- pctl_data %>%
  mutate(pred = predict(lin_mod, pctl_data),
         method = "percentile")

plot_data <- bind_rows(grid_data, pctl_data)

library(ggplot2)

ggplot(plot_data, aes(x = disp, y = pred)) +
  geom_point(alpha = .5, cex = 1) +
  facet_wrap(~ method)
# }

Run the code above in your browser using DataCamp Workspace