R/plot.check_distribution.R

In see: Model Visualisation Toolbox for 'easystats' and 'ggplot2'

#' Plot method for classifying the distribution of a model-family
#'
#' The `plot()` method for the `performance::check_distribution()`
#' function.
#'
#' @param panel Logical, if `TRUE`, plots are arranged as panels; else,
#'   single plots are returned.
#' @inheritParams data_plot
#' @inheritParams plot.see_bayesfactor_parameters
#'
#' @return A ggplot2-object.
#'
#' @examplesIf identical(Sys.getenv("NOT_CRAN"), "true") && require("randomForest")
#' library(performance)
#' m <<- lm(mpg ~ wt + cyl + gear + disp, data = mtcars)
#' result <- check_distribution(m)
#' result
#' plot(result)
#' @export
plot.see_check_distribution <- function(x, size_point = 2, panel = TRUE, ...) {
  model <- .retrieve_data(x)
  x <- x[-which(x$p_Residuals == 0 & x$p_Response == 0), ]

  dat <- data.frame(
    x = factor(c(x$Distribution, x$Distribution), levels = rev(sort(unique(x$Distribution)))),
    y = c(x$p_Response, x$p_Residuals),
    group = factor(
      c(rep("Response", length(x$p_Response)), rep("Residuals", length(x$p_Residuals))),
      levels = c("Response", "Residuals")
    ),
    stringsAsFactors = FALSE
  )

  # remove all zero-probabilities
  dat$y[dat$y == 0] <- NA

  # find max-range, add some tolerance
  max_y <- max(dat$y, na.rm = TRUE) * 1.1

  # default legend-position
  lp <- ifelse(isTRUE(panel), "right", "bottom")

  p1 <- ggplot(
    dat,
    aes(
      y = .data$x,
      x = .data$y,
      colour = .data$group
    )
  ) +
    geom_linerange(
      aes(xmin = 0, xmax = .data$y),
      position = position_dodge(0.4),
      linewidth = 0.8,
      na.rm = TRUE
    ) +
    geom_point(
      size = size_point,
      position = position_dodge(0.4),
      na.rm = TRUE
    ) +
    labs(
      y = NULL,
      x = NULL,
      fill = NULL,
      colour = NULL,
      title = "Predicted Distribution of Residuals and Response"
    ) +
    scale_x_continuous(
      labels = .percents,
      expand = c(0, 0),
      limits = c(0, max_y)
    ) +
    scale_color_material_d(reverse = TRUE) +
    guides(colour = guide_legend(reverse = TRUE)) +
    theme_lucid(legend.position = lp)

  dat1 <- as.data.frame(stats::density(stats::residuals(model)))
  dat2 <- data.frame(x = .factor_to_numeric(insight::get_response(model)))

  # this ensures that we have integers for response variables with values
  # much greater than 1 and with a large range, so we have proper bar plots
  if (diff(range(dat2$x)) > 10) dat2$x <- round(dat2$x)

  p2 <- ggplot(dat1, aes(x = .data$x, y = .data$y)) +
    geom_line(colour = "#2196F3") +
    labs(x = NULL, y = NULL, title = "Density of Residuals") +
    theme_lucid()

  # usually, we have an outline for the bars in the same color as the
  # theme background, however, for very thin bars, this results in "invisible"
  # bars, because the fill colour is not visible. For larger ranges, use fill
  # colour as color, too.

  if (abs(diff(range(dat2$x))) > 100) {
    bar_color <- "#f44336"
  } else {
    bar_color <- theme_lucid()$panel.background$fill
  }

  p3 <- ggplot(dat2, aes(x = .data$x)) +
    geom_histogram(
      fill = "#f44336", colour = bar_color,
      binwidth = sqrt(length(vars(.data$x))),
      na.rm = TRUE
    ) +
    labs(x = NULL, y = NULL, title = "Distribution of Response") +
    theme_lucid()

  if (panel) {
    insight::check_if_installed("patchwork")
    return(p1 / (p2 | p3) + patchwork::plot_layout(nrow = 2))
  } else {
    return(list(p1, p2, p3))
  }
}



#' @export
plot.see_check_distribution_numeric <- function(x,
                                                size_point = 2,
                                                panel = TRUE,
                                                ...) {
  vec <- .retrieve_data(x)
  x <- x[-which(x$p_Vector == 0), ]

  dat <- data.frame(
    x = factor(x$Distribution, levels = rev(sort(unique(x$Distribution)))),
    y = x$p_Vector,
    stringsAsFactors = FALSE
  )

  # remove all zero-probabilities
  dat$y[dat$y == 0] <- NA

  # find max-range, add some tolerance
  max_y <- max(dat$y, na.rm = TRUE) * 1.1

  # default legend-position
  lp <- ifelse(isTRUE(panel), "right", "bottom")

  p1 <- ggplot(dat, aes(y = .data$x, x = .data$y)) +
    geom_linerange(
      aes(xmin = 0, xmax = .data$y),
      position = position_dodge(0.4),
      linewidth = 0.8,
      na.rm = TRUE
    ) +
    geom_point(
      size = size_point,
      position = position_dodge(0.4),
      na.rm = TRUE
    ) +
    labs(
      y = NULL,
      x = NULL,
      fill = NULL,
      colour = NULL,
      title = "Predicted Distribution of Vector"
    ) +
    scale_x_continuous(
      labels = .percents,
      expand = c(0, 0),
      limits = c(0, max_y)
    ) +
    theme_lucid(legend.position = lp)

  dat1 <- as.data.frame(stats::density(vec))
  dat2 <- data.frame(x = vec, stringsAsFactors = FALSE)

  p2 <- ggplot(dat1, aes(x = .data$x, y = .data$y)) +
    geom_line() +
    labs(x = NULL, y = NULL, title = "Density of Vector") +
    theme_lucid()

  p3 <- ggplot(dat2, aes(x = .data$x)) +
    geom_histogram(
      colour = theme_lucid()$panel.background$fill,
      binwidth = sqrt(length(vars(.data$x))),
      na.rm = TRUE
    ) +
    labs(x = NULL, y = NULL, title = "Distribution of Vector") +
    theme_lucid()

  if (panel) {
    insight::check_if_installed("patchwork")
    return(p1 / (p2 | p3) + patchwork::plot_layout(nrow = 2L))
  } else {
    return(list(p1, p2, p3))
  }
}