R/ggplot_parallel_slopes.R

In moderndive/moderndive: Tidyverse-Friendly Introductory Linear Regression

globalVariables(c(
  "."
))

#' Plot parallel slopes model
#'
#' NOTE: This function is deprecated; please use [geom_parallel_slopes()]
#' instead. Output a visualization of linear regression when you have one numerical
#' and one categorical explanatory/predictor variable: a separate colored
#' regression line for each level of the categorical variable
#'
#' @inheritParams stats::lm
#' @param y Character string of outcome variable in `data`
#' @param num_x Character string of numerical explanatory/predictor variable in
#'   `data`
#' @param cat_x Character string of categorical explanatory/predictor variable
#'   in `data`
#' @param alpha Transparency of points
#' @seealso [geom_parallel_slopes()]
#' @return A [ggplot2::ggplot()] object.
#' @importFrom glue glue
#' @importFrom stats as.formula
#' @import ggplot2
#' @export
#'
#' @examples
#' \dontrun{
#' library(ggplot2)
#' library(dplyr)
#' library(moderndive)
#'
#' # log10() transformations
#' house_prices <- house_prices %>%
#'   mutate(
#'     log10_price = log10(price),
#'     log10_size = log10(sqft_living)
#'   )
#'
#' # Output parallel slopes model plot:
#' gg_parallel_slopes(
#'   y = "log10_price", num_x = "log10_size", cat_x = "condition",
#'   data = house_prices, alpha = 0.1
#' ) +
#'   labs(
#'     x = "log10 square feet living space", y = "log10 price in USD",
#'     title = "House prices in Seattle: Parallel slopes model"
#'   )
#'
#' # Compare with interaction model plot:
#' ggplot(house_prices, aes(x = log10_size, y = log10_price, col = condition)) +
#'   geom_point(alpha = 0.1) +
#'   geom_smooth(method = "lm", se = FALSE, size = 1) +
#'   labs(
#'     x = "log10 square feet living space", y = "log10 price in USD",
#'     title = "House prices in Seattle: Interaction model"
#'   )
#' }
gg_parallel_slopes <- function(y, num_x, cat_x, data, alpha = 1) {
  .Deprecated("geom_parallel_slopes()")
  # Define model formula and fitted/predicted value
  formula <- glue::glue(y, " ~ ", num_x, " + ", cat_x) %>%
    as.formula()
  y_hat <- paste(y, "_hat", sep = "")

  # Get fitted values for parallel slopes model
  model_points <- data %>%
    lm(formula, data = .) %>%
    get_regression_points()

  plot <-
    # Plot scatterplot of points
    ggplot(data, aes_string(x = num_x, y = y, col = cat_x)) +
    geom_point(alpha = alpha) +
    # Add parallel slopes lines. Note the data and aes(y = ) override:
    geom_line(data = model_points, aes_string(y = y_hat), size = 1, show.legend = FALSE) +
    guides(colour = guide_legend(override.aes = list(alpha = 1)))

  message_wrap("Plotting parallel slopes models is now much easier using the 
  `geom_parallel_slopes()` function. We suggest you use `geom_parallel_slopes()` instead
  of `gg_parallel_slopes()`; read the help file by running `?geom_parallel_slopes` to 
               learn how.")

  return(plot)
}