ggh4x

The ggh4x package is an extension for ggplot2 that provides utility functions for advanced plot tweaking that fall outside the standard 'grammar of graphics'.

Key capabilities include:

Facet Customization

• Extended Facets: Additional options for axis labelling and placement.
• Nested Facets: Strips that can span multiple panels.
• Manual Facets: Custom layouts.
• Custom Strips: More types of strips for facet labels.
• Per-panel Position Scales: Adjusting scales on a per-panel basis.
• Panel Sizing: Varying the size of panels without being limited by global aspect.ratio or fixed coordinates.

Statistical Layers

• Theoretical Densities: Overlaying densities computed with the fitdistrplus package.
• Rolling Kernels: Drawing trend lines using a rolling kernel.
• Group-wise Transformations: Transforming x and y positions in a group-wise manner.
• Run-length Encoding: Calculating run-length encodings of data.

To use multiple colour scales in a single plot, you must first map data to 'alternative aesthetics' (custom names) within your geom_* layers. scale_colour_multi then maps these custom aesthetics to specific colour palettes.

Note: ggplot2 will issue warnings about 'unknown aesthetics' when you use custom names like SW or PL; this is expected behavior.

Arguments provided as lists to scale_colour_multi are passed on to individual scales.

library(ggh4x)
library(scales)

df <- transform(iris, Nester = ifelse(Species == "setosa", "Short Leaves", "Long Leaves"))

g <- ggplot(df, aes(Sepal.Width, Sepal.Length)) +
  theme_classic() +
  theme(strip.background = element_blank())

# 1. Map data to alternative aesthetics (e.g., SW, PL, PW)
g <- g + 
  geom_point(aes(SW = Sepal.Width), data = ~ subset(., Species == "setosa")) +
  geom_point(aes(PL = Petal.Length), data = ~ subset(., Species == "versicolor")) +
  geom_point(aes(PW = Petal.Width), data = ~ subset(., Species == "virginica"))

# 2. Apply the multi-colour scale
g <- g +
  scale_colour_multi(
    aesthetics = c("SW", "PL", "PW"),
    name = list("Blue", "Pink", "Orange"),
    colours = list(
      brewer_pal(palette = "YlGnBu")(6),
      brewer_pal(palette = "RdPu")(6),
      brewer_pal(palette = "YlOrRd")(6)
    ),
    guide = guide_colorbar(barheight = unit(50, "pt"))
  )

g

The facetted_pos_scales() function allows you to specify unique scales (labels, breaks, limits, transformations, or guides) for each individual facet panel.

Usage Requirements:

• The list of scales must follow the order of the facets (assuming scales are set to 'free').
• This function must be called after the facets have been added to the ggplot object.

Important Limitation: Because scale transformations are applied after stat calculations when using facetted_pos_scales(), it can cause issues with layers using non-identity statistics (like geom_density()). For such layers, it is recommended to pre-transform the data in the aes() mapping.

# Define a list of scales for each facet
scales <- list(
  scale_x_reverse(),
  scale_x_continuous(labels = scales::dollar, minor_breaks = c(2.5, 4.5)),
  scale_x_continuous(breaks = c(2.945, 6), limits = c(0, 10), guide = "axis_minor")
)

# Apply to the plot after adding facets
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  facet_wrap(vars(drv), scales = "free_x") +
  facetted_pos_scales(x = scales)

The force_panelsizes() function allows you to set absolute or relative sizes for facet rows and columns. This is useful for creating layouts where panels have specific dimensions (e.g., placing a density plot next to a scatter plot).

Key features:

• Overriding: These settings overrule coordinates' aspect ratios, theme aspect ratios, and space = "free" facet arguments.
• Relative vs Absolute:
  • By default, rows and columns are relative within themselves.
  • Setting respect = TRUE makes the relative units consistent across both rows and columns.
  • You can use grid::unit() to specify absolute sizes.
• Requirement: Must be added after any facets.

# Set columns to relative sizes 1 and 0.2, and rows to 0.5, respecting aspect ratio
g <- ggplot(faithful) +
  geom_point(aes(waiting, eruptions), data = cbind(faithful, facet = "Points")) +
  geom_density(aes(y = eruptions), data = cbind(faithful, facet = "Density")) +
  facet_grid(~ facet, scales = "free_x") +
  force_panelsizes(cols = c(1, 0.2), 
                   rows = c(0.5), 
                   respect = TRUE)

The facet_nested() function allows for hierarchical facet strips where outer strips can span multiple inner strips belonging to the same category. This is useful for representing hierarchical relationships in data.

Key features:

• Nesting lines: You can indicate hierarchy using lines instead of background strips by setting nest_line = TRUE. The appearance is controlled by the theme element ggh4x.facet.nestline.
• Data flexibility: Unlike facet_grid(), facet_nested() does not require the input data to contain all facet variables.
• Strip ordering: When strips are placed at the bottom (using switch = "x"), facet_nested() rearranges them so inner strips are closest to the panels and spanning strips are furthest away.

# Basic nested facets
ggplot(df, aes(Sepal.Width, Sepal.Length)) +
  geom_point() +
  facet_nested(~ Nester + Species)

# Nested facets with lines and custom theme
ggplot(df, aes(Sepal.Width, Sepal.Length)) +
  geom_point() +
  facet_nested(~ Nester + Species, nest_line = TRUE) +
  theme(strip.background = element_blank(),
        ggh4x.facet.nestline = element_line(colour = "blue"))

The facet_grid2() function extends ggplot2::facet_grid() with enhanced axis control and 'independent' scales.

Key arguments:

• axes: Controls axis placement at inner facets (e.g., axes = "all").
• remove_labels: Controls which axis labels to omit (e.g., remove_labels = "y").
• independent: Allows scales to be free within rows and columns of the layout, rather than just between rows or columns. For example, independent = "x" allows scales to be free within columns.

Note on constraints: You cannot use independent scales and space = "free" for the same dimension. However, you can combine them if they apply to different dimensions (e.g., independent = "y" and space = "free_x").

p <- ggplot(mpg, aes(displ, hwy, colour = as.factor(cyl))) + geom_point()

# Use independent x scales and show axes on all inner facets
p + facet_grid2(vars(drv), vars(year), axes = "all", remove_labels = "y")

# Use independent x scales with free x space (not allowed) vs independent y with free x space (allowed)
p + facet_grid2(vars(drv), vars(year), scales = "free", independent = "y", space = "free_x")