ggstackplot

2025-01-31

# load the package
library(ggstackplot)

This vignette explores the various features of the ggstackplot package.

Main Arguments

x and y arguments

Vertical stack

Select variables to make a stack. The selection order translates to the order with which the plots are stacked. Any valid tidyselect selection and/or renaming are supported.

# select any number of variables to make the stack
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec, drat)
  )


# the selection order translates into stack order
mtcars |> 
  ggstackplot(
    x = mpg, y = c(drat, wt, qsec)
  )


# use any valid tidyselect selection syntax
mtcars |> 
  ggstackplot(
    x = mpg, y = c(4, "carb", starts_with("d"))
  )


# use any valid tidyselect renaming syntax to rename stack panels
mtcars |> 
  ggstackplot(
    x = c(`mpg [units]` = mpg), 
    y = c(`weight [tons]` = wt, `speed` = qsec, drat)
  )

Horizontal stack

Select multiple x variables to stack:

# all examples shown in this document work the same way for a horizontal
# stack, simply switch out the x and y assignments
mtcars |> 
  ggstackplot(
    y = mpg, x = c(wt, qsec, drat)
  )

palette argument

Set individual plot colors by providing an RColorBrewer palette. Color definition applies to the color and fill aesthetics as well as the actual axis colors.

# use the Set1 RColorBrewer palette
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    palette = "Set1"
  )

# likewise for the horizontal stack version
mtcars |> 
  ggstackplot(
    y = mpg, x = c(wt, qsec),
    palette = "Set1"
  )

color argument

Alternatively, set colors manually by supplying a character vector of colors:

# select any specific colors for each plot
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8")
  )

remove_na argument

This removes NA values so that lines are not interrupted. When remove_na is set to FALSE, breaks in lines may appear due to NA values.

library(dplyr)

# default (NAs are removed so lines are not interrupted)
mtcars |> 
  add_row(mpg = 22, wt = 5, qsec = NA) |>
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8")
  )


# explicit `remove_na` = FALSE
mtcars |> 
  add_row(mpg = 22, wt = 5, qsec = NA) |>
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    remove_na = FALSE
  )

both_axes argument

When both_axes = TRUE , the stacked variable axes are duplicated on both sides of each stacked plot.

# Vertical stackplot
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    both_axes = TRUE
  )


# Horizontal stackplot
mtcars |> 
  ggstackplot(
    y = mpg, x = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    both_axes = TRUE
  )

alternate_axes argument

When alternate_axes = FALSE , the axes for the multiple variables are kept on the same side of the facets. The default behavior alternates these axes left/right or top/bottom.

# axes do not alternate:
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    alternate_axes = FALSE
  )


# Horizontal version
mtcars |> 
  ggstackplot(
    y = mpg, x = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    alternate_axes = FALSE
  )

switch_axes argument

Determines whether to switch the stacked axes. Not switching means that for vertical stacks the plot at the bottom has the y-axis always on the left side; and for horizontal stacks that the plot on the left has the x-axis on top. Setting switch_axes = TRUE}, leads to the opposite. If alternate_axes = TRUE this essentially switches the order with which the axes alternate (e.g., right/left/right vs. left/right/left). Note that if both_axes = TRUE, neither the switch_axes nor alternate_axes parameter has any effect.

# stacked axis starts on the right
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    switch_axes = TRUE
  )


# or for the horizontal version, stacked axis
# starts on the bottom
mtcars |> 
  ggstackplot(
    y = mpg, x = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    switch_axes = TRUE
  )


# and in combination with alternate_axes = FALSE
# all axes on the right
mtcars |> 
  ggstackplot(
    x = mpg, y = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    alternate_axes = FALSE,
    switch_axes = TRUE
  )


# or all axes on the top
mtcars |> 
  ggstackplot(
    y = mpg, x = c(wt, qsec),
    color = c("#E41A1C", "#377EB8"),
    alternate_axes = FALSE,
    switch_axes = TRUE
  )

overlap argument

Overlap determines the grid overlap between the multiple stacked plots. 1 corresponds to fully overlapping (similar to having a ggplot sec_axis enabled) while 0 does not overlap at all.

# define any overlap between 0 and 1
mtcars |> 
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 0.3
  )


# full overlap
mtcars |> 
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 1
  )

Different overlaps

Multiple overlap arguments can be supplied with a numeric vector of numbers between 0 and 1, where each element in the vector corresponds to the overlap between the n and n+1th overlap value. For example, for a plot with four stacked panels: qsec, drat, wt, hp, a vector of overlap = c(1, 0, 1) indicates that between the first 2 elements (qsec and drat) there is full overlap. Between drat and wt there is no overlap (0). Between wt and hp there is full overlap.

# different overlap between stack panels
mtcars |> 
  ggstackplot(
    x = mpg, 
    y = c(qsec, drat, wt, hp),
    color = c("#E41A1C", "#377EB8", "#4DAF4A", "#984EA3"),
    overlap = c(1, 0, 1)
  )


# and the horizontal version
mtcars |> 
  ggstackplot(
    y = mpg, 
    x = c(qsec, drat, wt, hp),
    color = c("#E41A1C", "#377EB8", "#4DAF4A", "#984EA3"),
    overlap = c(1, 0, 1)
  )

shared_axis_size argument

The size of the shared axis determines the size of any shared axes relative to the grid size of the original ggplot. The size of the shared axis often needs to be adjusted depending on which aspect ratio is intended. It is defined as fraction of a full panel, between 0 and 1.

mtcars |> 
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 1,
    # can be only 10% of a plot size as we're overlapping plots
    shared_axis_size = 1
  )

simplify_shared_axis argument

Sometimes it’s better just to keep the shared axis on each panel. This produces something akin to a facet_wrap() or cowplot::plot_grid().

mtcars |> 
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    simplify_shared_axis = FALSE
  )


# also goes well with changing `both_axes`, `switch_axes` and/or `alternate_axes`
mtcars |> 
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    simplify_shared_axis = FALSE,
    alternate_axes = FALSE
  )

The template argument

This is the most important argument. It defines which ggplot to use as the template for all plots in the stack. This can be an actual plot (just the data will be replaced) or a ggplot that doesn’t have data associated yet. The possibilities are pretty much endless. Just make sure to always add the theme_stacked_plot() base theme (you can modify it more from there on). A few examples below:

Theme modifications

Add any modification to the overlying theme as you see fit.

Here, template allows the user to define that a ggplot() will serve as the base, with geom_line as the primary geom. Then, theme_stackplot() is applied and custom theme() options are set.

library(ggplot2)

# increase y axis text size
mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    template = 
      ggplot() + 
      geom_line() +
      theme_stackplot() +
      theme(
        axis.title.y = element_text(size = 20),
        axis.text.y = element_text(size = 16)
      )
  )


# increase the panel margins
mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    template = 
      ggplot() + 
      geom_line() +
      theme_stackplot() +
      theme(
        # increase left margin to 20% and top/bottom margins to 10%
        plot.margin = margin(l = 0.2, t = 0.1, b = 0.1, unit = "npc")
      )
  )

Grid modifications

Modifying the panel.grid argument can create gridlines for both the stacked variable axes and the shared axis. This can get a bit cluttered in a plot where overlap = 1.

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 1,
    template = ggplot() +
      geom_line(data = function(df) filter(df, .yvar == "qsec")) +
      geom_point(data = function(df) filter(df, .yvar == "drat")) +
      theme_stackplot() +
      theme(
        panel.grid.major = element_line(
          color = "lightgray", 
          linewidth = 0.8)
      )
  )

But, this can look reasonable if there is no overlap of the stacked plats, and/or if the lines are made inconspicuous:

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 0,
    template = ggplot() +
      geom_line(data = function(df) filter(df, .yvar == "qsec")) +
      geom_point(data = function(df) filter(df, .yvar == "drat")) +
      theme_stackplot() +
      theme(
        panel.grid.major = element_line(
          color = "lightgray", 
          linetype = "dotted", 
          linewidth = 0.5)
      )
  )

Other themes

You aren’t bound to our theme’s aesthetic choices :), you can always add another theme or theme modifications on top of theme_stackplot()! Here we add the classic theme_bw() to get those nice clean gridlines back, as well as a panel border.

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 0,
    template = ggplot() +
      geom_line(data = function(df) filter(df, .yvar == "qsec")) +
      geom_point(data = function(df) filter(df, .yvar == "drat")) +
      theme_stackplot() +
      theme_bw() # give us that good theme!
  )

Custom geom data

It is possible to use different geoms for different stacked panels. Here, we use both lines and points. These geoms are defined in the template argument.

# use different geoms for different panels
# you can refer to y-stack panel variables with `.yvar` and x-stack panel variables with `.xvar`
mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 1,
    template = ggplot() +
      geom_line(data = function(df) filter(df, .yvar == "qsec")) +
      geom_point(data = function(df) filter(df, .yvar == "drat")) +
      theme_stackplot()
  )

Different plot elements

One can also change the geoms in the default theme. Here we use geom_path() instead of geom_line() in a horizontal stack. This is a very common use case because geom_line() connects the data points by increasing x-axis which is not always what we want (for example in oceanographic depth plots where we want to connect the data points by increasing y-axis value).

# horizontal stack with default (geom_line())
mtcars |>
  ggstackplot(
    y = mpg, x = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    template = 
      ggplot() + 
      geom_point() +
      geom_line() + # default in template
      theme_stackplot() 
  )

# the following is the exact same data but using a
# horizontal stack with "depth-profile" like geom_path()
mtcars |>
  # arrange data by the y-axis
  arrange(mpg) |>
  ggstackplot(
    y = mpg, x = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    template = 
      ggplot() + 
      geom_point() +
      geom_path() + # plots data in order
      theme_stackplot() 
  )

Additional plot elements

One can also add additional plot elements just as with a normal ggplot. Here we add a vertical line that is shared across all stacked plots:

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    overlap = 0.2, 
    template = 
      ggplot() + 
      geom_vline(xintercept = 20, linewidth = 4, color = "gray80") +
      geom_line() +
      theme_stackplot() 
  )

Axis modifications

Sometimes secondary axes will still be desired, especially if that axis is a transformation of an existing one. For example, here, we create a square root mpg axis that is plotted against the mpg axis. All this can also be defined in the template argument by adding a scale_x_continuous argument, just as you would in a normal ggplot.

# add a secondary x axis
mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    both_axes = TRUE, overlap = 0.1, 
    template = 
      ggplot() + 
      geom_line() +
      scale_x_continuous(
        # change axis name
        name = "this is my mpg axis",
        # this can be the same with dup_axis() or as here have a transformed axis
        sec.axis = sec_axis(
          transform = sqrt, 
          name = expression(sqrt(mpg)), 
          breaks = scales::pretty_breaks(5)
        )
      ) +
      theme_stackplot() 
  )

Similarly, transformation axes can be introduced such as e.g. a log axis.

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    both_axes = TRUE, overlap = 0.1,
    template = 
      ggplot() + 
      geom_line() +
      scale_x_log10("this is my log10 mpg axis") +
      theme_stackplot() 
  )

Additional aesthetics

Aesthetics are also defined in the template argument. Remember, the only parameters that are defined in the stackplot are (i) the shared axis (in this case, mpg ), (ii) the axes to be stacked, in this case y = c(wt, qsec, drat), (iii) any ggstackplot-specific arguments. All ggplot arguments and aesthetics are assigned in the template argument.

# add aesthetics to the plot
mtcars |>
  ggstackplot(
    x = mpg,  y = c(wt, qsec, drat),
    alternate_axes = FALSE,
    template = 
      ggplot() +
      aes(color = factor(cyl), linetype = factor(cyl), shape = factor(cyl)) +
      geom_line() +
      geom_point(size = 3) +
      theme_stackplot() 
  )

The add argument

For even more specific plot refinements, the add argument provides an easy way to add ggplot components to specific panels in the stack plot. A few examples below:

Custom geoms

Similar to the example custom geom data the add argument can also be used to add specific geoms only to specific panels.

This takes the form of a list() where each item in the list is of the form: panel_name = panel_addition where panel_name is the panel-specific variable and panel_addition is the item to add (+) to that panel. add also allows the user to make additions by index (e.g., first panel, second panel, third panel, etc.).

Here, we add a geom_line to the qsec panel and a geom_rect rectangle to the drat panel by defining these panels in the list().

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    template = ggplot() + theme_stackplot(),
    # add:
    add = list(
      # panel by name
      qsec = geom_line(), 
      drat = geom_rect(
        xmin = 20, xmax = 25, ymin = 3.2, ymax = 4.2, fill = "gray90") + 
        geom_point()
    )
  )

Custom themes

Similarly, custom theme options can be added to specific panels. Here, we add by panel index:

mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    # define ggplot template options
    template = 
      ggplot() + 
      geom_line() + 
      theme_stackplot(),
    # define panel-specific additions
    add = list(
      # panel by index
      # first panel:
      geom_point() + theme(
        axis.title.y = element_text(size = 30)),
      # second panel:
      theme(
        panel.grid.major.y = element_line(
          color = "lightgray", 
          size = 0.2))
    )
  )

Custom axes

The add argument also allows the definition of custom axes. This is particularly useful if applying functions from the scales package.

# particularly useful is also the possibility to modify individual scales
mtcars |>
  ggstackplot(
    x = mpg, y = c(qsec, drat),
    color = c("#E41A1C", "#377EB8"),
    template = ggplot() + geom_line() + theme_stackplot(),
    add = list(
      # modify the axis for the second plot
      drat = 
        scale_y_continuous("$$ drat",  labels = scales::label_dollar()) + 
        expand_limits(y = 0) +
        theme(axis.title.y = element_text(size = 30))
    )
  )

Legend positioning

Another example of theme modification is the use of the `add` argument to specify legend positioning.

mtcars |>
  ggstackplot(
    x = mpg,  y = c(wt, qsec, drat),
    color = c("#E41A1C", "#377EB8", "#4DAF4A"),
    template = 
      ggplot() + aes(linetype = factor(vs)) +
      geom_line() + theme_stackplot(),
    # switch legend position for middle plot
    add = list(qsec = theme(legend.position = "left"))
  )


mtcars |>
  ggstackplot(
    x = mpg,  y = c(wt, qsec, drat),
    color = c("#E41A1C", "#377EB8", "#4DAF4A"),
    template = 
      ggplot() + 
      aes(linetype = factor(vs)) +
      geom_line() + 
      theme_stackplot() +
      # remove the legends, then...
      theme(legend.position = "none"), 
    # ... re-include the middle panel legend on the plot
    # with some additional styling
    add = list(
      qsec = 
        theme(
          # define legend relative position in x,y:
          legend.position = c(0.2, 0.9), 
          # other legend stylistic changes:
          legend.title = element_text(size = 20),
          legend.text = element_text(size = 16),
          legend.background = element_rect(
            color = "black", fill = "gray90", linewidth = 0.5),
          legend.key = element_blank(),
          legend.direction = "horizontal"
        ) +
        labs(linetype = "VS")
    )
  )

Putting it all together

# example from the README with economics data bundled with ggplot2
ggplot2::economics |>
  ggstackplot(
    # define shared x axis
    x = date, 
    # define the stacked y axes
    y = c(pce, pop, psavert, unemploy),
    # pick the RColorBrewer Dark2 palette (good color contrast)
    palette = "Dark2",
    # overlay the pce & pop plots (1), then make a full break (0) to the once
    # again overlaye psavert & unemploy plots (1)
    overlap = c(1, 0, 1),
    # switch axes so unemploy and psavert are on the side where they are 
    # highest, respectively - not doing this here by changing the order of y
    # because we want pop and unemploy on the same side
    switch_axes = TRUE,
    # make shared axis space a bit smaller
    shared_axis_size = 0.15,
    # provide a base plot with shared graphics eelements among all plots
    template = 
      # it's a ggplot
      ggplot() +
      # use a line plot for all
      geom_line() +
      # we want the default stackplot theme
      theme_stackplot() +
      # add custom theme modifications, such as text size
      theme(text = element_text(size = 14)) +
      # make the shared axis a date axis
      scale_x_date("year") +
      # include y=0 for all plots to contextualize data better
      expand_limits(y = 0),
    # add plot specific elements
    add = 
      list(
        pce = 
          # show pce in trillions of dollars
          scale_y_continuous(
            "personal consumption expenditures", 
            # always keep the secondary axis duplicated so ggstackplot can
            # manage axis placement for you
            sec.axis = dup_axis(),
            # labeling function for the dollar units
            labels = function(x) sprintf("$%.1f T", x/1000),
          ),
        pop = 
          # show population in millions
          scale_y_continuous(
            "population", sec.axis = dup_axis(),
            labels = function(x) sprintf("%.0f M", x/1000)
          ),
        psavert = 
          # savings is in %
          scale_y_continuous(
            "personal savings rate", sec.axis = dup_axis(),
            labels = function(x) paste0(x, "%"),
          ) +
          # show data points in addition to line
          geom_point(),
        unemploy = 
          # unemploy in millions
          scale_y_continuous(
            "unemployed persons", sec.axis = dup_axis(),
            labels = function(x) sprintf("%.0f M", x/1000)
          ) +
          # show data points in addition to line
          geom_point()
      )
  )

Advanced

Instead of calling ggstackplot() to make a plot, you can also use prepare_stackplot() and assemble_stackplot() to separate the two main steps of making a ggstackplot. prepare_stackplot() provides a tibble with all the plot components that can be modified directly in the tibble if so desired before assembling the plot with assemble_stackplot(). Usuallyt this is not necessary because the combination of the template and add parameters in ggstackplot() provides the same kind of flexibility as modifying plot elements in the plot tibble.

# prep plot
plot_prep <- 
  mtcars |> 
  prepare_stackplot(
    x = mpg, y = c(wt, qsec),
    palette = "Set1"
  )

# show plot tibble
plot_prep
#> # A tibble: 2 × 6
#>   .var  config           data               plot   theme   add   
#>   <chr> <list>           <list>             <list> <list>  <list>
#> 1 wt    <tibble [1 × 9]> <tibble [32 × 11]> <gg>   <theme> <NULL>
#> 2 qsec  <tibble [1 × 9]> <tibble [32 × 11]> <gg>   <theme> <NULL>

# modify plot tibble
plot_prep$plot[[2]] <- ggplot(mtcars) + aes(mpg, drat) + geom_point()
plot_prep$theme[[2]] <- theme_bw()

# assemble stackplot
plot_prep |> assemble_stackplot()