Data visualization cookbook for the Center for Rural Innovation
Source:vignettes/cookbook.Rmd
cookbook.RmdSetup
Load fonts into your environment
# For easily loading Lato, Bitter, and TT Hoves (TT Hoves needs to be installed locally)
# you can use the following utility function
cori.charts::load_fonts()
# Otherwise, I recommend loading fonts using the sysfonts library
# Loads Lato from the google font repository and adds it to sysfonts
sysfonts::font_add_google("Lato")
# Ensures that any newly opened graphics devices will use showtext to draw text
showtext::showtext_auto()
# Sets default density per inch for exports
showtext::showtext_opts(dpi = 300)Load your data
We will be using a dataset which gives the population with an income below the poverty level in the last 12 months for a subset of RIN communities.
data(cori_poverty)
dplyr::glimpse(cori_poverty)
#> Rows: 10
#> Columns: 8
#> $ rin_community <chr> "Berkshire", "Emporia", "Marquette",…
#> $ state_abbr <chr> "MA", "KS", "MI", "KY", "OH", "OK", …
#> $ county <chr> "Berkshire", "Crawford", "Marquette"…
#> $ geoid_co <chr> "25003", "20037", "26103", "21195", …
#> $ estimate_pop_2014 <dbl> 130064, 39277, 67535, 64380, 78520, …
#> $ estimate_pop_2019 <dbl> 126425, 38968, 66686, 59497, 76040, …
#> $ estimate_below_poverty_level_2014 <dbl> 15746, 8102, 10597, 15238, 18284, 83…
#> $ estimate_below_poverty_level_2019 <dbl> 13063, 7513, 9833, 14545, 16330, 731…Now we are ready to begin charting!
Chart Recipes
Horizontal Bar Chart
cori_poverty %>%
ggplot2::ggplot(
aes(
estimate_below_poverty_level_2019,
# Make RIN communities appear in descending order
# based upon their population below the poverty level
reorder(rin_community, estimate_below_poverty_level_2019)
)
) +
# Set the bar color the the CORI "Emerald" color
ggplot2::geom_col(fill = cori_colors["Emerald"]) +
# Add data labels to the bars
geom_text(
aes(
# You can adjust the scales function depending on the desired
# number format (e.g., percent, dollar, etc.)
label = scales::number(
estimate_below_poverty_level_2019,
# accuracy determines what number to round to (e.g., accuracy = 0.01 will show 2 decimal places)
accuracy = 1,
# big.mark determines the character used between every 3 digits to separate thousands
big.mark = ","
)
),
fontface = "bold",
# Provide spacing between the data label and the bar position
hjust = -.2,
# Data labels need to have their font family explicitly set to "Lato"
family = "Lato"
) +
ggplot2::scale_x_continuous(
# labels determines whether tick labels are shown
labels = NULL,
# You can provide an expansion multiplier to the axis to ensure that
# data labels will have enough space
expand = expansion(mult = c(0, .25))
) +
# Call the horizontal bar theme to pull in default CORI theming
theme_cori_horizontal_bars() +
# Override any defaults styles using the ggplot2::theme() function AFTER
# calling theme_cori_horizontal_bars()
ggplot2::theme(
# Set title fonts to Lato, as TT Hoves is unavailable
# Where possible, use TT Hoves for titles (default)
plot.title = element_text(family = "Lato"),
plot.subtitle = element_text(family = "Lato"),
# Remove x gridlines
panel.grid.major.x = element_blank()
) +
# Provide Title, subtitle, etc.
ggplot2::labs(
title = "Population living below the poverty level",
subtitle = "For select RIN communities (2019)",
y = NULL,
x = NULL,
caption = "Source: ACS 5-year estimates (2019)"
)
Grouped Bar Chart
Pivot the data into a long format for easy plotting
grouped_bar_data <- cori_poverty %>%
mutate(
percent_below_poverty_2014 = estimate_below_poverty_level_2014 / estimate_pop_2014,
percent_below_poverty_2019 = estimate_below_poverty_level_2019 / estimate_pop_2019
) %>%
pivot_longer(
contains("percent"),
names_to = "year",
values_to = "percent_below_poverty"
) %>%
mutate(
year = str_remove(year, "percent_below_poverty_"),
rin_community = reorder(rin_community, percent_below_poverty)
)
glimpse(grouped_bar_data)
#> Rows: 20
#> Columns: 10
#> $ rin_community <fct> Berkshire, Berkshire, Emporia, Empor…
#> $ state_abbr <chr> "MA", "MA", "KS", "KS", "MI", "MI", …
#> $ county <chr> "Berkshire", "Berkshire", "Crawford"…
#> $ geoid_co <chr> "25003", "25003", "20037", "20037", …
#> $ estimate_pop_2014 <dbl> 130064, 130064, 39277, 39277, 67535,…
#> $ estimate_pop_2019 <dbl> 126425, 126425, 38968, 38968, 66686,…
#> $ estimate_below_poverty_level_2014 <dbl> 15746, 15746, 8102, 8102, 10597, 105…
#> $ estimate_below_poverty_level_2019 <dbl> 13063, 13063, 7513, 7513, 9833, 9833…
#> $ year <chr> "2014", "2019", "2014", "2019", "201…
#> $ percent_below_poverty <dbl> 0.12106348, 0.10332608, 0.20627848, …With data labels
grouped_bar_data %>%
ggplot(aes(percent_below_poverty, rin_community, fill = year)) +
geom_col(position = "dodge") +
# Add in data labels
geom_text(
aes(label = scales::percent(percent_below_poverty, accuracy = 1)),
# Need to add a position value to ensure that the data labels
# are aligned with their appropriate bar
position = position_dodge2(width = 0.9, reverse = FALSE),
hjust = -.2,
family = "Lato",
fontface = "bold"
) +
scale_fill_cori(palette = "ctg2tlpu", reverse = TRUE) +
scale_x_continuous(
# Axis labels are redundant to the data labels,
# so we don't display them by setting labels = NULL
labels = NULL,
expand = expansion(mult = c(0, .1))
) +
labs(
title = "Share of population below poverty level",
subtitle = "For select CORI communities",
x = NULL,
y = NULL,
caption = "Source: U.S. Census Bureau"
) +
theme_cori_horizontal_bars() +
theme(
# Set title fonts to Lato, as TT Hoves is unavailable
# Where possible, use TT Hoves for titles (default)
plot.title = element_text(family = "Lato"),
plot.subtitle = element_text(family = "Lato")
)
Without data labels
grouped_bar_data %>%
ggplot(aes(percent_below_poverty, rin_community, fill = year)) +
geom_col(position = "dodge") +
scale_fill_cori(palette = "ctg2buor", reverse = TRUE) +
scale_x_continuous(
# Add back axis labels to provide context
# because we aren't displaying data labels
labels = scales::label_percent(accuracy = 1),
expand = expansion(mult = c(0, .1))
) +
labs(
title = "Share of population below poverty level",
subtitle = "For select CORI communities",
x = NULL,
y = NULL,
caption = "Source: U.S. Census Bureau"
) +
theme_cori_horizontal_bars() +
theme(
# Set title fonts to Lato, as TT Hoves is unavailable
# Where possible, use TT Hoves for titles (default)
plot.title = element_text(family = "Lato"),
plot.subtitle = element_text(family = "Lato")
)
Bullet Chart
# Create a column for the width of the two bars using dplyr:: mutate()
#
# Generally, when comparing time periods, the wide bar is the
# earlier time period and the skinny bar is the later time period
bullet_chart_data <- grouped_bar_data %>%
dplyr::mutate(
width = ifelse(year == "2014", 0.75, 0.5)
)
glimpse(bullet_chart_data)
#> Rows: 20
#> Columns: 11
#> $ rin_community <fct> Berkshire, Berkshire, Emporia, Empor…
#> $ state_abbr <chr> "MA", "MA", "KS", "KS", "MI", "MI", …
#> $ county <chr> "Berkshire", "Berkshire", "Crawford"…
#> $ geoid_co <chr> "25003", "25003", "20037", "20037", …
#> $ estimate_pop_2014 <dbl> 130064, 130064, 39277, 39277, 67535,…
#> $ estimate_pop_2019 <dbl> 126425, 126425, 38968, 38968, 66686,…
#> $ estimate_below_poverty_level_2014 <dbl> 15746, 15746, 8102, 8102, 10597, 105…
#> $ estimate_below_poverty_level_2019 <dbl> 13063, 13063, 7513, 7513, 9833, 9833…
#> $ year <chr> "2014", "2019", "2014", "2019", "201…
#> $ percent_below_poverty <dbl> 0.12106348, 0.10332608, 0.20627848, …
#> $ width <dbl> 0.75, 0.50, 0.75, 0.50, 0.75, 0.50, …
bullet_chart_data %>%
ggplot(aes(x = percent_below_poverty, y = rin_community, fill = year)) +
# Determine the width using the "width" column we created above
geom_col(width = bullet_chart_data$width) +
# Add data labels for the later time period only
geom_text(
data = dplyr::filter(bullet_chart_data, year == "2019"),
aes(x = percent_below_poverty, label = scales::percent(percent_below_poverty, accuracy = 1)),
hjust = 1.3,
family = "Lato",
color = "white",
fontface = "bold"
) +
scale_fill_cori(palette = "ctg2buor", reverse = TRUE) +
scale_x_continuous(
labels = scales::label_percent(accuracy = 1),
expand = expansion(mult = c(0, .1))
) +
labs(
title = "Share of population below poverty level",
subtitle = "For select CORI communities",
x = NULL,
y = NULL,
caption = "Source: ACS 5-year estimates"
) +
theme_cori_horizontal_bars() +
theme(
# Set title fonts to Lato, as TT Hoves is unavailable
# Where possible, use TT Hoves for titles (default)
plot.title = element_text(family = "Lato"),
plot.subtitle = element_text(family = "Lato")
)
Line Charts
data("cori_education")
# Clean up the labels for the chart
line_chart_data <- cori_education %>%
dplyr::mutate(
# Remove underscores and convert to title case
education_clean = stringr::str_to_title(stringr::str_replace_all(education, "_", " "))
)
# Prep work so that the legend labels can be ordered based upon their final data point
#
# Filter to the final data point and pull the order
latest_date <- line_chart_data %>% pull(date) %>% max()
factor_order <- line_chart_data %>%
dplyr::filter(date == latest_date) %>%
dplyr::arrange(desc(percent_working_remotely)) %>%
pull(education_clean)
# Update the column based upon the preferred order
line_chart_data <- line_chart_data %>%
dplyr::mutate(education_clean = factor(education_clean, levels = factor_order))
glimpse(line_chart_data)
#> Rows: 80
#> Columns: 4
#> $ date <date> 2020-05-01, 2020-05-01, 2020-05-01, 2020-05-…
#> $ education <fct> less_than_high_school, high_school_no_college…
#> $ percent_working_remotely <dbl> 0.05, 0.15, 0.25, 0.60, 0.05, 0.13, 0.22, 0.5…
#> $ education_clean <fct> Less Than High School, High School No College…Simple line chart with right-hand legend
# When using line charts, you need to
# update the geom defaults before plotting
update_cori_geom_defaults()
line_chart_data %>%
ggplot(
aes(date, percent_working_remotely, color = education_clean)) +
geom_line() +
scale_color_cori(palette = "ctg4mid", guide = guide_legend(reverse = TRUE)) +
scale_x_date(
date_breaks = "6 months",
date_labels = "%b '%y",
expand = expansion(mult = c(.01,.01))
) +
scale_y_continuous(
labels = scales::label_percent(accuracy = 1),
limits = c(0, .65),
expand = expansion(mult = c(0, .1))
) +
theme_cori() +
ggplot2::theme(
# CORI themes defaults to a horizontal legend below the title,
# but with long names it can be easier to display the legend on the right
legend.position = "right",
legend.direction = "vertical",
# Add ticks to the x-axis
axis.ticks.x = element_line(color = "#d0d2ce"),
axis.ticks.length = unit(4, 'pt'),
plot.title = element_text(family = "Lato"),
plot.subtitle = element_text(family = "Lato")
) +
# In order for a legend on the right to render properly,
# you need to specify that byrow = TRUE
guides(color = guide_legend(byrow = TRUE)) +
labs(title = "Percent of workers able to work from home",
subtitle = "By education level (May 2020 - December 2021)",
x = NULL,
y = NULL,
caption = "Source: Bureau of Labor Statistics")
If your lines are spaced further apart, you can use a secondary axis to
directly label the line endpoints
# Let's filter to just two lines to simplify the chart
direct_label_data <- line_chart_data %>%
dplyr::filter(
education == "bachelors_or_higher" | education == "some_college_or_associates"
)
# Determine the value and label at the most recent date for each line
line_labels <- direct_label_data %>%
dplyr::filter(date == latest_date) %>%
dplyr::arrange(desc(percent_working_remotely)) %>%
pull(education_clean)
line_values <- direct_label_data %>%
dplyr::filter(date == latest_date) %>%
dplyr::arrange(desc(percent_working_remotely)) %>%
pull(percent_working_remotely)
direct_label_data %>%
ggplot(
aes(date, percent_working_remotely, color = education_clean)) +
geom_line() +
scale_color_cori(palette = "ctg2tlpu", guide = guide_legend(reverse = TRUE)) +
scale_x_date(
date_breaks = "6 months",
date_labels = "%b '%y",
expand = expansion(mult = c(.01,0))
) +
scale_y_continuous(
labels = scales::label_percent(accuracy = 1),
limits = c(0, .65),
expand = expansion(mult = c(0, .1)),
# Add the direct labels as a second axis
sec.axis = sec_axis(
trans = ~.*1,
breaks = line_values,
labels = line_labels
)
) +
theme_cori() +
ggplot2::theme(
# Remove the legend
legend.position = "none",
# Add ticks to the x-axis
axis.ticks.x = element_line(color = "#d0d2ce"),
axis.ticks.length = unit(8, 'pt'),
# Add margin to separate tick and text
axis.text.x = element_text(margin = margin(t = 2)),
# Specify title font. Use TT Hoves if possible
plot.title = element_text(family = "Lato"),
plot.subtitle = element_text(family = "Lato"),
# Adjust the space between the line and the line label
axis.ticks.length.y.right = unit(1, "pt")
) +
labs(
title = "Percent of workers able to work from home",
subtitle = "By education level (May 2020 - December 2021)",
x = NULL,
y = NULL,
caption = "Source: Bureau of Labor Statistics"
)
#> Warning: The `trans` argument of `sec_axis()` is deprecated as of ggplot2 3.5.0.
#> ℹ Please use the `transform` argument instead.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.
Color selection
Available colors
Commonly used colors from our branding
guidelines can be easily accessed using cori.charts. To
specify a color, simply access the cori_colors list with
the name of the color (e.g., cori_colors["Emerald"])
Color palettes
Several default color palettes are also provided. If you call
scale_color_cori or scale_fill_cori without
specifying the palette argument it defaults to Emerald and
Bright Mint:
This default palette is not always the best choice, depending on your data’s format.
If you are comparing two categories (e.g., Rural vs. Nonrural), I recommend using either the Mid Teal and Mid Purple palette
or the Mid Blue and Mid Orange palette.
If you want to show four categories, I recommend using the 4 category Mid Color palette.
While there is a 7 category palette, I would think carefully before going beyond four categories in any graphic.
