plot_loadings: Plot PCA component loadings from a recipe or workflow

In tidymodels/learntidymodels: Learn 'tidymodels' with Interactive 'learnr' Primers

Description

A plot of the component loadings for each original column of the data is produced.

Usage

plot_loadings(x, ...)

## S3 method for class 'recipe'
plot_loadings(x, ..., id = NULL, type = "pca")

## S3 method for class 'workflow'
plot_loadings(x, ..., id = NULL)

Arguments

x	A prepped recipe or fitted workflow that uses a recipe. The recipe must have used at least one recipes::step_pca() (or recipes::step_pls()).
...	An optional series of conditional statements used to filter the PCA data before plotting. See Details below.
id	A single numeric or character value that is used to pick the step with the PCA results. If a single recipes::step_pca() (or recipes::step_pls()) was used, this argument is ignored. Note: if used, id must be named.
type	A character value ("pca" or "pls") for the type of step to use.

Value

A ggplot object.

Examples

library(recipes)
library(parsnip)
library(workflows)
library(ggplot2)

data("Chicago", package = "modeldata")

theme_set(theme_minimal())

## -----------------------------------------------------------------------------

train_pca <-
  recipe(ridership ~ ., data = Chicago %>% dplyr::select(1:21)) %>%
  step_center(all_predictors()) %>%
  step_scale(all_predictors()) %>%
  step_pca(all_predictors())

# or when used in a workflow
lm_workflow <-
  workflow() %>%
  add_model(linear_reg() %>% set_engine("lm")) %>%
  add_recipe(train_pca)

## -----------------------------------------------------------------------------

train_pca <- prep(train_pca)

plot_loadings(train_pca, component_number <= 3)

plot_loadings(train_pca, component_number <= 3, value > 0)

## -----------------------------------------------------------------------------

lm_workflow <- lm_workflow %>% fit(data = Chicago)

plot_loadings(lm_workflow, component_number <= 3)

tidymodels/learntidymodels documentation built on Dec. 14, 2021, 5:12 p.m.