Summary bar graphs

Graph info

Should I use this graph?

This graph requires:

✅ a numeric (continuous) variable

✅ a categorical variable

Description

Summary bar graphs display the sum (or total) of a numerical variable across the levels of a second categorical variable. Color is used to make comparisons and distinguish between groups (or levels) of the categorical variable.

In ggplot2, we can create summary bar graphs with geom_bar().

Getting set up

Packages

PACKAGES:

Install packages.

Code

install.packages("palmerpenguins")
library(palmerpenguins) 
library(ggplot2)

Data

DATA:

Artwork by @allison_horst

Remove the missing values from body_mass_g and island in the palmerpenguins::penguins data and convert body mass in grams to kilograms (body_mass_kg).

We’ll also reduce the number of columns in the penguins data for clarity.

Code

peng_sum_col <- palmerpenguins::penguins |> 
    dplyr::select(body_mass_g, island) |> 
    tidyr::drop_na() |> 
    # divide the mass value by 1000
    dplyr::mutate(
        body_mass_kg = body_mass_g / 1000
    )
dplyr::glimpse(peng_sum_col)

Rows: 342
Columns: 3
$ body_mass_g  <int> 3750, 3800, 3250, 3450, 3650, 3625, 4675, 3475, 4250, 330…
$ island       <fct> Torgersen, Torgersen, Torgersen, Torgersen, Torgersen, To…
$ body_mass_kg <dbl> 3.750, 3.800, 3.250, 3.450, 3.650, 3.625, 4.675, 3.475, 4…

The grammar

Code

CODE:

Create labels with labs()

Initialize the graph with ggplot() and provide data

Map island to x and body_mass_kg to y

Inside the aes() of geom_col(), map island to fill

Outside the aes() of geom_col(), remove the legend with show.legend = FALSE

Code

labs_sum_col <- labs(
    title = "Total Penguin Mass",
    subtitle = "How many kilograms of penguin per Island?",
    x = "Island",
    y = "Total Penguin Body Mass (kg)")
ggp2_sum_col <- ggplot(data = peng_sum_col,
    aes(x = island,
        y = body_mass_kg)) +
    geom_col(aes(fill = island), 
        show.legend = FALSE)  
ggp2_sum_col + 
    labs_sum_col

Graph

GRAPH:

More Info

Note that we didn’t have to write any code to calculate the total body_mass_g (displayed on the y axis) by island.

That’s because ggplot2 does this for us!

Summary

SUMMARY:

If we pass a categorical variable to the x (like island) and a continuous variable to y (like body_mass_kg), geom_col() will calculate the sum() of y by levels of x

We can see the underlying summary of budget using dplyr’s group_by() and summarise() functions.

Code

palmerpenguins::penguins |> 
    dplyr::select(body_mass_g, island) |> 
    tidyr::drop_na() |> 
    # divide the mass value by 1000
    dplyr::mutate(
        body_mass_kg = body_mass_g / 1000
    ) |> 
    dplyr::group_by(island) |>
    dplyr::summarise(
       `Total Penguin Body Mass (kg)` = sum(body_mass_kg)) |>
    dplyr::ungroup() |> 
    dplyr::select(`Island` = island, 
        `Total Penguin Body Mass (kg)`)

Island	Total Penguin Body Mass (kg)
Biscoe	787.575
Dream	460.400
Torgersen	189.025

Identity vs. Count

STATS:

The geom_bar() geom will also create grouped bar graphs, but we have to switch the stat argument to "identity".

Code

ggplot(data = peng_sum_col,
    aes(x = island,
        y = body_mass_kg)) +
    geom_col(aes(fill = island), 
        show.legend = FALSE,
        stat = "identity")  +
    labs_sum_col

geom_bar() vs. geom_col():

geom_bar() will map a categorical variable to the x or y and display counts for the discrete levels (see stat_count() for more info)

geom_col() will map both x and y aesthetics, and is used when we want to display numerical (quantitative) values across the levels of a categorical variable. geom_col() assumes these values have been created in their own column (see stat_identity() for more info)