Visualizing Data (in R)An opinionated style guide

June Choe2021-02-101 / 73

<a href='https://xkcd.com/1945/'>https://xkcd.com/1945</a>

2 / 73

Motivation3 / 73

Motivation

Why bother improving the quality of your plot?

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible
Increases the chance of your work being shared

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible
Increases the chance of your work being shared
Shows respect for your audience

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible
Increases the chance of your work being shared
Shows respect for your audience

Goal of this presentation

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible
Increases the chance of your work being shared
Shows respect for your audience

Goal of this presentation

Share practical, language-agnostic tips for improving readability & accessibility

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible
Increases the chance of your work being shared
Shows respect for your audience

Goal of this presentation

Share practical, language-agnostic tips for improving readability & accessibility
Walk through implementations of these ideas (w/ R code)

3 / 73

Motivation

Why bother improving the quality of your plot?

Helps people remember your results better
Makes your findings look more credible
Increases the chance of your work being shared
Shows respect for your audience

Goal of this presentation

Share practical, language-agnostic tips for improving readability & accessibility
Walk through implementations of these ideas (w/ R code)
Demonstrate that the design just as important as the content

3 / 73

Outline4 / 73

Outline

Four principles for explanatory data visualization

4 / 73

Outline

Four principles for explanatory data visualization

Make your text readable

4 / 73

Outline

Four principles for explanatory data visualization

Make your text readable
Be generous about margins and spacing

4 / 73

Outline

Four principles for explanatory data visualization

Make your text readable
Be generous about margins and spacing
Make the legend clear but subtle

4 / 73

Outline

Four principles for explanatory data visualization

Make your text readable
Be generous about margins and spacing
Make the legend clear but subtle
Make color easy on your reader's eyes

4 / 73

Outline

Four principles for explanatory data visualization

Make your text readable
Be generous about margins and spacing
Make the legend clear but subtle
Make color easy on your reader's eyes

Showcases/tricks

Minimal designs
Plot composition
Animations
Interactivity

4 / 73

Preliminaries

The {ggplot2} ecosystem in R:

Easy to use and highly customizable
Tons of free resources for learning
Widely used in academia
Lots of extensions

5 / 73

Preliminaries

The {ggplot2} ecosystem in R:

Easy to use and highly customizable
Tons of free resources for learning
Widely used in academia
Lots of extensions

library(extrafont)
library(colorspace)
library(patchwork)
library(ggtext)
library(scales)
# library(gghighlight)
# library(ggforce)
# library(ggrepel)
# library(gganimate)

6 / 73

The layered Grammar of Graphics

A framework is less rules, not more

Video Tutorial

7 / 73

# Presidential election results from 1932-2016
state_election_votes

  # A tibble: 1,097 x 3
     State       demVote  Year
     <chr>         <dbl> <dbl>
   1 Alabama       0.85   1932
   2 Arizona       0.67   1932
   3 Arkansas      0.86   1932
   4 California    0.580  1932
   5 Colorado      0.55   1932
   6 Connecticut   0.47   1932
   7 Delaware      0.48   1932
   8 Florida       0.74   1932
   9 Georgia       0.92   1932
  10 Idaho         0.59   1932
  # ... with 1,087 more rows
7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  )

  # A tibble: 66 x 3
     State        demVote  Year
     <chr>          <dbl> <dbl>
   1 California     0.580  1932
   2 Illinois       0.55   1932
   3 Pennsylvania   0.45   1932
   4 California     0.67   1936
   5 Illinois       0.580  1936
   6 Pennsylvania   0.570  1936
   7 California     0.570  1940
   8 Illinois       0.51   1940
   9 Pennsylvania   0.53   1940
  10 California     0.56   1944
  # ... with 56 more rows
7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot()

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_point()

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_point() +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_point() +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  ) +
  scale_x_continuous(
    breaks = pretty_breaks(n= 5)
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_point() +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  ) +
  scale_x_continuous(
    breaks = pretty_breaks(n= 5)
  ) +
  labs(
    y = "Democrat Votes",
    title = "Go Vote!"
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_point() +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  ) +
  scale_x_continuous(
    breaks = pretty_breaks(n= 5)
  ) +
  labs(
    y = "Democrat Votes",
    title = "Go Vote!"
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_point() +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  ) +
  scale_x_continuous(
    breaks = pretty_breaks(n= 5)
  ) +
  labs(
    y = "Democrat Votes",
    title = "Go Vote!"
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_line() +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  ) +
  scale_x_continuous(
    breaks = pretty_breaks(n= 5)
  ) +
  labs(
    y = "Democrat Votes",
    title = "Go Vote!"
  )

7 / 73

# Presidential election results from 1932-2016
state_election_votes %>%
  filter(State %in%
    c("Pennsylvania",
    "Illinois",
    "California")
  ) %>%
  ggplot() +
  aes(
    x = Year,
    y = demVote,
    color = State
  ) +
  geom_line(size = 1) +
  scale_y_continuous(
    labels = percent_format(accuracy = 1)
  ) +
  scale_x_continuous(
    breaks = pretty_breaks(n= 5)
  ) +
  labs(
    y = "Democrat Votes",
    title = "Go Vote!"
  )

7 / 73

Yay a plot!

8 / 73

Yay a plot... ?

Areas of improvement

Text is small and narrow
Plot elements are squished together
Difficult to focus on individual colors
Legend is off to the side on its own

9 / 73

1. Text10 / 73

1. Text

Many different ways to style text:

11 / 73

1. Text

Many different ways to style text:

font size

11 / 73

1. Text

Many different ways to style text:

font size
Font face (bold, italic, small caps)

11 / 73

1. Text

Many different ways to style text:

font size
Font face (bold, italic, small caps)
Font family ( Times New Roman, Calibri, Arial, Hi friends )

11 / 73

1. Text

Many different ways to style text:

font size
Font face (bold, italic, small caps)
Font family ( Times New Roman, Calibri, Arial, Hi friends )
Font color, letter spacing, angle, weight, etc.

11 / 73

1. Text

Many different ways to style text:

font size
Font face (bold, italic, small caps)
Font family ( Times New Roman, Calibri, Arial, Hi friends )
Font color, letter spacing, angle, weight, etc.

Often times you'll likely only have access to a limited set:

11 / 73

1. Text

Many different ways to style text:

font size
Font face (bold, italic, small caps)
Font family ( Times New Roman, Calibri, Arial, Hi friends )
Font color, letter spacing, angle, weight, etc.

Often times you'll likely only have access to a limited set:

Font size, font family, and limited set of font styles

11 / 73

1. Text

Many different ways to style text:

font size
Font face (bold, italic, small caps)
Font family ( Times New Roman, Calibri, Arial, Hi friends )
Font color, letter spacing, angle, weight, etc.

Often times you'll likely only have access to a limited set:

Font size, font family, and limited set of font styles
But you can get pretty far with these!

11 / 73

state_election_plot

11 / 73

state_election_plot +
  theme(text = element_text(size = 15))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot")

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year")

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL)

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state")

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!")

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Open Sans")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!") +
  theme(plot.subtitle = element_text(face = "italic"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Adelle")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!") +
  theme(plot.subtitle = element_text(face = "italic"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Bitter")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!") +
  theme(plot.subtitle = element_text(face = "italic"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Montserrat")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!") +
  theme(plot.subtitle = element_text(face = "italic"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "Montserrat Medium")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!") +
  theme(plot.subtitle = element_text(face = "italic"))

11 / 73

state_election_plot +
  theme(text = element_text(size = 15)) +
  theme(plot.title = element_text(size = 20)) +
  theme(text = element_text(family = "xkcd")) +
  theme(plot.title = element_text(family = "Roboto Slab")) +
  theme(plot.title.position = "plot") +
  labs(x = "Election Year") +
  labs(y = NULL) +
  labs(title = "Percent of democrat votes by state") +
  labs(subtitle = "We're a swing state! Go vote!") +
  theme(plot.subtitle = element_text(face = "italic"))

11 / 73

1. Text (End!)

state_election_plot +
  theme(
    text = element_text(size = 15, family = "Open Sans"),
    plot.title = element_text(size = 20, family = "Roboto Slab"),
    plot.title.position = "plot",
    plot.subtitle = element_text(face = "italic"),
    axis.text = element_text(size = 14)
  ) +
  labs(
    x = "Election Year",
    y = NULL,
    title = "Percent of democrat votes by state",
    subtitle = "We're a swing state! Go vote!"
  )

Save our progress!

state_election_plot_A

12 / 73

1. Text (Before-After)

13 / 73

2. Margin & Spacing14 / 73

2. Margins & Spacing

15 / 73

2. Margins & Spacing

16 / 73

2. Margins & Spacing

17 / 73

2. Margins & Spacing

Margin and spacing considerations:

18 / 73

2. Margins & Spacing

Margin and spacing considerations:

Margins around plot make it easier to embed figures

18 / 73

2. Margins & Spacing

Margin and spacing considerations:

Margins around plot make it easier to embed figures
Spacing between text elements improves readability

18 / 73

2. Margins & Spacing

Margin and spacing considerations:

Margins around plot make it easier to embed figures
Spacing between text elements improves readability
Disproportionate white (empty) space can be distracting

18 / 73

2. Margins & Spacing

Margin and spacing considerations:

Margins around plot make it easier to embed figures
Spacing between text elements improves readability
Disproportionate white (empty) space can be distracting
Padding inside the panel is essential for a sense of "completeness"

18 / 73

2. Margins & Spacing

Margin and spacing considerations:

Margins around plot make it easier to embed figures
Spacing between text elements improves readability
Disproportionate white (empty) space can be distracting
Padding inside the panel is essential for a sense of "completeness"

Know your margin/spacing elements in {ggplot2}!

margin(t = 0, r = 0, b = 0, l = 0, unit = c("pt", "mm", "cm", "in"))
hjust, vjust, and lineheight arguments in element_text()
expand argument with expansion() in scale_*() and coord_*() layers

18 / 73

state_election_plot_A

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  ) +
  theme(
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm"))
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  ) +
  theme(
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm"))
  ) +
  theme(
    axis.title.x = element_text(margin = margin(t = .3, unit = "cm"))
  )

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  ) +
  theme(
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm"))
  ) +
  theme(
    axis.title.x = element_text(margin = margin(t = .3, unit = "cm"))
  ) +
  coord_cartesian(expand = FALSE)

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  ) +
  theme(
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm"))
  ) +
  theme(
    axis.title.x = element_text(margin = margin(t = .3, unit = "cm"))
  ) +
  scale_x_continuous(expand = expansion(add = 3))

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  ) +
  theme(
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm"))
  ) +
  theme(
    axis.title.x = element_text(margin = margin(t = .3, unit = "cm"))
  ) +
  scale_x_continuous(expand = expansion(add = c(2, 3)))

18 / 73

state_election_plot_A +
  theme(
    plot.background = element_rect(color = 'black')
  ) +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm")
  ) +
  theme(
    plot.title = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm"))
  ) +
  theme(
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm"))
  ) +
  theme(
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm"))
  ) +
  theme(
    axis.title.x = element_text(margin = margin(t = .3, unit = "cm"))
  ) +
  scale_x_continuous(expand = expansion(add = c(2, 3)), breaks = pretty_breaks(5))

18 / 73

2. Margins & Spacing (End!)

state_election_plot_A +
  theme(
    plot.margin = margin(1, .8, .8, .8, "cm"),
    plot.title = element_text(margin = margin(b = .3, unit = "cm")),
    plot.subtitle = element_text(margin = margin(b = .3, unit = "cm")),
    axis.text.x = element_text(margin = margin(t = .2, unit = "cm")),
    axis.text.y = element_text(margin = margin(r = .1, unit = "cm")),
    axis.title.x = element_text(margin = margin(t = .3, unit = "cm"))
  ) +
  scale_x_continuous(
    expand = expansion(add = c(2, 3)),
    breaks = pretty_breaks(5)
  )

Save our progress!

state_election_plot_B

19 / 73

2. Margins & Spacing (Before-After)

20 / 73

2. Margins & Spacing (Before-After)

21 / 73

3. Legends22 / 73

3. Legends23 / 73

3. Legends

Legends are really hard:

23 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place

23 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it

23 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it
But sometimes we don't have the luxury of doing so

23 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it
But sometimes we don't have the luxury of doing so

Alternative: consider labeling the data directly

23 / 73

24 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it
But sometimes we don't have the luxury of doing so

Alternative: consider labeling the data directly

25 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it
But sometimes we don't have the luxury of doing so

Alternative: consider labeling the data directly

But if you must have a legend, make it so that the reader is naturally guided to it

25 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it
But sometimes we don't have the luxury of doing so

Alternative: consider labeling the data directly

But if you must have a legend, make it so that the reader is naturally guided to it

Positioned at the top-left or top-center of the plot

25 / 73

3. Legends

Legends are really hard:

They contain important info, but difficult to make them not look out of place
It's better not to have a legend, if you can get away with it
But sometimes we don't have the luxury of doing so

Alternative: consider labeling the data directly

But if you must have a legend, make it so that the reader is naturally guided to it

Positioned at the top-left or top-center of the plot
Blends smoothly into the rest of the plot (NOT make them stand out!)

25 / 73

state_election_plot_B

25 / 73

state_election_plot_B +
  theme(
    legend.key = element_rect(color = NA, fill = NA),
  )

25 / 73

state_election_plot_B +
  theme(
    legend.key = element_rect(color = NA, fill = NA),
  ) +
  theme(
    legend.position = c(.45, .93)
  )

25 / 73

state_election_plot_B +
  theme(
    legend.key = element_rect(color = NA, fill = NA),
  ) +
  theme(
    legend.position = c(.45, .93)
  ) +
  theme(
    legend.direction = "horizontal"
  )

25 / 73

state_election_plot_B +
  theme(
    legend.key = element_rect(color = NA, fill = NA),
  ) +
  theme(
    legend.position = c(.45, .93)
  ) +
  theme(
    legend.direction = "horizontal"
  ) +
  theme(
    legend.background = element_rect(fill = "grey92")
  )

25 / 73

state_election_plot_B +
  theme(
    legend.key = element_rect(color = NA, fill = NA),
  ) +
  theme(
    legend.position = c(.45, .93)
  ) +
  theme(
    legend.direction = "horizontal"
  ) +
  theme(
    legend.background = element_rect(fill = "grey92")
  ) +
  scale_y_continuous(
    expand = expansion(add = c(0.02, .05)),
    labels = percent_format(accuracy = 1)
  )

25 / 73

state_election_plot_B +
  theme(
    legend.key = element_rect(color = NA, fill = NA),
  ) +
  theme(
    legend.position = c(.45, .93)
  ) +
  theme(
    legend.direction = "horizontal"
  ) +
  theme(
    legend.background = element_rect(fill = "grey92")
  ) +
  scale_y_continuous(
    expand = expansion(add = c(0.02, .05)),
    labels = percent_format(accuracy = 1)
  ) +
  labs(color = NULL)

25 / 73

3. Legends (End!)

state_election_plot_B +
  theme(
    legend.key = element_rect(fill = NA),
    legend.position = c(.45, .93),
    legend.direction = "horizontal",
    legend.background = element_rect(fill = "grey92")
  ) + 
  scale_y_continuous(
    expand = expansion(add = c(0.02, .05)),
    labels = percent_format(accuracy = 1)
  ) +
  labs(color = NULL)

Save our progress!

state_election_plot_C

26 / 73

3. Legends (Before-After)

27 / 73

4. Color28 / 73

4. Color29 / 73

4. Color

Colors are a double-edged sword

29 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.

29 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.
Color space is complex; no "all in one" solution (RGB/hex, HSV, HCL)

29 / 73

chroma::show_col(rainbow(6))

30 / 73

specplot(rainbow(6))

31 / 73

HSV

32 / 73

HSV

HCL

33 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.

Color space is complex - no "all in one" solution (RGB, hex, HSV, HCL)

34 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.

Color space is complex - no "all in one" solution (RGB, hex, HSV, HCL)

If you must use colors, DO:

34 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.

Color space is complex - no "all in one" solution (RGB, hex, HSV, HCL)

If you must use colors, DO:

Avoid pure colors (no random sampling from the rainbow!)

34 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.

Color space is complex - no "all in one" solution (RGB, hex, HSV, HCL)

If you must use colors, DO:

Avoid pure colors (no random sampling from the rainbow!)
Contrast colors in more than one dimension

34 / 73

source: https://blog.datawrapper.de/beautifulcolors/

35 / 73

source: https://blog.datawrapper.de/beautifulcolors/

36 / 73

Correlation matrix example

corr_left <- ggcorrplot::ggcorrplot(cor(mtcars)) + scale_fill_gradient2(low = "blue", mid = "white", high = "red", limits = c(-1, 1), guide = guide_colorbar(title = NULL, barheight = unit(15, 'lines')))
corr_right <- ggcorrplot::ggcorrplot(cor(mtcars)) + scale_fill_continuous_diverging(palette = "Blue-Red 3", limits = c(-1, 1), guide = guide_colorbar(title = NULL, barheight = unit(15, 'lines')))
patchwork::wrap_plots(corr_left, corr_right, nrow = 1)

37 / 73

Correlation matrix example

corr_desaturated_left <- cowplot::ggdraw(colorblindr::edit_colors(corr_left, desaturate))
corr_desaturated_right <- cowplot::ggdraw(colorblindr::edit_colors(corr_right, desaturate))
patchwork::wrap_plots(corr_desaturated_left, corr_desaturated_right, nrow = 1)

38 / 73

Correlation matrix example

palette_left <- scales::div_gradient_pal(low = "blue", mid = "white", high = "red")(seq(0, 1, length.out = 7))
palette_right <- colorspace::diverge_hcl(7, palette = "Blue-Red 3")

39 / 73

4. Color

Colors are a double-edged sword

Perception can vary widely depending on reader, medium, culture, etc.

Color space is complex; no "all in one" solution (RGB, hex, HSV, HCL)

If you must, DO:

Avoid pure colors (no random sampling from the rainbow!)

Contrast colors in more than one dimension

If you aren't sure, use pre-made palettes or play around with online color tools

40 / 73

state_election_plot_C

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA))

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA))

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line())

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  )

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 27,   0.62,    0.95)),
      hex(HSV( 16,   0.49,    0.55)),
      hex(HSV(338,   0.96,    0.36))
    )
  )

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 27,   0.62,    0.95)),
      hex(HSV( 16,   0.49,    0.55)),
      hex(HSV(338,   0.96,    0.36))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,   0.62,       1)),
      hex(HSV( 16,   0.49,       1)),
      hex(HSV(338,   0.96,       1))
    )
  )

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 27,   0.62,    0.95)),
      hex(HSV( 16,   0.49,    0.55)),
      hex(HSV(338,   0.96,    0.36))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,   0.62,       1)),
      hex(HSV( 16,   0.49,       1)),
      hex(HSV(338,   0.96,       1))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,      1,    0.95)),
      hex(HSV( 16,      1,    0.55)),
      hex(HSV(338,      1,    0.36))
    )
  )

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 27,   0.62,    0.95)),
      hex(HSV( 16,   0.49,    0.55)),
      hex(HSV(338,   0.96,    0.36))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,   0.62,       1)),
      hex(HSV( 16,   0.49,       1)),
      hex(HSV(338,   0.96,       1))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,      1,    0.95)),
      hex(HSV( 16,      1,    0.55)),
      hex(HSV(338,      1,    0.36))
    )
  ) +
  geom_line(size = 1.5)

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 27,   0.62,    0.95)),
      hex(HSV( 16,   0.49,    0.55)),
      hex(HSV(338,   0.96,    0.36))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,   0.62,       1)),
      hex(HSV( 16,   0.49,       1)),
      hex(HSV(338,   0.96,       1))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,      1,    0.95)),
      hex(HSV( 16,      1,    0.55)),
      hex(HSV(338,      1,    0.36))
    )
  ) +
  geom_line(size = 1.5) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  )

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 27,   0.62,    0.95)),
      hex(HSV( 16,   0.49,    0.55)),
      hex(HSV(338,   0.96,    0.36))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,   0.62,       1)),
      hex(HSV( 16,   0.49,       1)),
      hex(HSV(338,   0.96,       1))
    )
  ) +
  scale_color_manual(
    values = c(
      hex(HSV( 33,      1,    0.95)),
      hex(HSV( 16,      1,    0.55)),
      hex(HSV(338,      1,    0.36))
    )
  ) +
  geom_line(size = 1.5) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  ) +
  scale_color_manual(values = desaturate(c("#F2A05C", "#8C5A48", "#5C0424")))

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  ggthemes::scale_color_colorblind()

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  ggsci::scale_color_simpsons()

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  ghibli::scale_color_ghibli_d("SpiritedMedium")

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  scale_color_manual(values = wesanderson::wes_palette("Moonrise2", 3, "discrete"))

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  scale_color_manual(values = LaCroixColoR::lacroix_palette("PeachPear", 3, "discrete"))

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  list(scale_color_manual(values = c("#65657F", "#855F4C", "#556760")), annotation_raster(bernie, xmin = 1986, xmax = 2016, ymin = .4, ymax = .65))

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  palettetown::scale_color_poke(pokemon = "pikachu")

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  theme(plot.background = element_rect(fill = "#ACACAC"), panel.grid = element_blank()) +
  palettetown::scale_color_poke(pokemon = "pikachu")

40 / 73

state_election_plot_C +
  theme(panel.background = element_rect(fill = NA)) +
  theme(legend.background = element_rect(fill = NA)) +
  theme(axis.line = element_line()) +
  geom_line(size = 1.5) +
  theme(plot.background = element_rect(fill = "#ACACAC"), panel.grid = element_blank()) +
  scale_color_manual(values = sequential_hcl(3, palette = "ag_Sunset"))

40 / 73

Plot

saturated_sunset <- chroma::as.hsv(sequential_hcl(3, palette = "ag_Sunset"))[, 'h'] %>% 
  map_chr(~ hex(HSV(H = .x, S = 1, V = 1)))
state_election_plot_C +
  theme(
    axis.line = element_line(color = "coral"),
    axis.text = element_text(color = "white"),
    legend.background = element_rect(fill = "#0F0D1A", color = "white", size = 0.5, linetype = "dashed"),
    panel.background = element_rect(fill = NA),
    panel.grid.major.x = element_blank(),
    panel.grid.major.y = element_line(color = "red", linetype = "dotted"),
    panel.grid.minor = element_blank(),
    plot.background = element_rect(fill = "#0F0D1A"),
    plot.subtitle = element_text(family = "VCR OSD Mono"),
    plot.title = element_text(family = "Blade Runner Movie Font"),
    text = element_text(color = "white", family = "SF Alien Encounters")
  ) +
  lemon::coord_capped_cart(left = "top") +
  lemon::geom_pointline(size = 2, distance = unit(2, 'pt'), key_glyph = "timeseries") +
  scale_color_manual(
    values = saturated_sunset,
    guide = guide_legend(override.aes = list(size = 1))
  )

41 / 73

state_election_plot_C

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  )

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  )

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  )

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  )

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  guides(
    color = guide_none()
  )

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  guides(
    color = guide_none()
  ) +
  geom_line(size = 1.5)

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  guides(
    color = guide_none()
  ) +
  geom_line(size = 1.5) +
  theme(plot.subtitle = ggtext::element_markdown())

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  guides(
    color = guide_none()
  ) +
  geom_line(size = 1.5) +
  theme(plot.subtitle = ggtext::element_markdown()) +
  labs(subtitle = "<span style='color:#2b5e82'>Pennsylvania</span> is a swing state! Go vote!")

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  guides(
    color = guide_none()
  ) +
  geom_line(size = 1.5) +
  theme(plot.subtitle = ggtext::element_markdown()) +
  labs(subtitle = "<strong style='color:#2b5e82'>Pennsylvania</strong> is a swing state! Go vote!")

41 / 73

state_election_plot_C +
  theme(
    panel.background = element_rect(fill = NA)
  ) +
  theme(
    legend.background = element_rect(fill = NA)
  ) +
  theme(
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  guides(
    color = guide_none()
  ) +
  geom_line(size = 1.5) +
  theme(plot.subtitle = ggtext::element_markdown()) +
  labs(subtitle = "<strong style='color:#0E4369'>Pennsylvania</strong> is a swing state! Go vote!")

41 / 73

4. Colors v.1 (End!)

state_election_plot_C +
  geom_line(size = 1.5) +
  theme(
    panel.background = element_rect(fill = NA),
    legend.background = element_rect(fill = NA),
    axis.line = element_line()
  ) +
  scale_color_manual(
    values = c("#F2A05C", "#8C5A48", "#5C0424")
  )

Save our progress!

state_election_plot_D1

42 / 73

4. Colors v.2 (End!)

state_election_plot_C +
  geom_line(size = 1.5) +
  theme(
    panel.background = element_rect(fill = NA),
    legend.background = element_rect(fill = NA),
    axis.line = element_line(),
    plot.subtitle = element_markdown()
  ) +
  guides(color = guide_none()) +
  scale_color_manual(values = c("#e1e1e1", "#e1e1e1", "#2b5e82")) +
  labs(
    subtitle = "<strong style='color:#0E4369'>Pennsylvania</strong>
                is a swing state! Go vote!"
  )

Save our progress!

state_election_plot_D2

43 / 73

4. Colors (Before-After)

44 / 73

4. Colors (Before-After)

45 / 73

state_election_plot

45 / 73

state_election_plot +
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  )

45 / 73

state_election_plot +
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  ) +
  guides(color = guide_none())

45 / 73

state_election_plot +
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  ) +
  guides(color = guide_none()) +
  geom_line(size = 1.5)

45 / 73

state_election_plot +
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  ) +
  guides(color = guide_none()) +
  geom_line(size = 1.5) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  )

45 / 73

state_election_plot +
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  ) +
  guides(color = guide_none()) +
  geom_line(size = 1.5) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  labs(
    y = NULL,
    x = "Election Year",
    title = "Percent of democrat votes by state",
    subtitle = "<strong style='color:#0E4369'>Pennsylvania</strong> is a swing state! Go vote!"
  )

45 / 73

state_election_plot +
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  ) +
  guides(color = guide_none()) +
  geom_line(size = 1.5) +
  scale_color_manual(
    values = c("#e1e1e1", "#e1e1e1", "#2b5e82")
  ) +
  labs(
    y = NULL,
    x = "Election Year",
    title = "Percent of democrat votes by state",
    subtitle = "<strong style='color:#0E4369'>Pennsylvania</strong> is a swing state! Go vote!"
  ) +
  theme(
    plot.margin = margin(.8, 1, .7, .8, "cm"),
    plot.title = element_text(
      family = "Roboto Slab",
      size = 20,
      margin = margin(b = .3, unit = "cm")
    ),
    plot.title.position = "plot",
    plot.subtitle = element_markdown(
      margin = margin(b = .3, unit = "cm")
    ),
    axis.title.x = element_text(
      margin = margin(t = .5, unit = "cm")
    )
  )

45 / 73

5(?). Plot Quality46 / 73

5(?). Plot quality: Resolution

47 / 73

Point-and-click method in RStudio

48 / 73

5(?). Plot quality: Scaling

49 / 73

Stretching in RStudio

50 / 73

5(?). Plot quality: Solution

Use the ggsave() function (only works for ggplot2 plots):

Automatically configures sensible defaults
Makes your figures fully reproducible!

ggsave("highres.png", p, width = 10, height = 6.2, units = "in", dpi = 300) # png format
ggsave("highres.pdf", p, width = 10, height = 6.2, units = "in", device = cairo_pdf) # pdf format w/ Cairo

51 / 73

5(?). Plot quality: Solution

Use the ggsave() function (only works for ggplot2 plots):

Automatically configures sensible defaults
Makes your figures fully reproducible!

ggsave("highres.png", p, width = 10, height = 6.2, units = "in", dpi = 300) # png format
ggsave("highres.pdf", p, width = 10, height = 6.2, units = "in", device = cairo_pdf) # pdf format w/ Cairo

Use better-quality, OS-independent graphic devices (works for any figure):

library(ragg) # R interface to the AGG device - https://github.com/r-lib/ragg
# Step 1: initialize device
agg_png("highres_agg.png", width = 10, height = 6.2, units = "in")
# Step 2: plot
plot(p)
# Step 3: close device
invisible(dev.off())

51 / 73

ggsave() png

52 / 73

ggsave() pdf

53 / 73

ragg png

54 / 73

5(?). Plot quality: Takeaways

1. Plot objects are not images

object.size(state_election_plot_D2)

  34032 bytes

fs::dir_info(path = "img", regexp = "point-and-click.png")$size

  43.6K

2. The conversion of plot to image is not neutral

Depends on dimensions, resolution, type of graphic device/operating system, etc.
Exert full control over the process to avoid any surprises at the end!

Useful reference: https://bookdown.org/rdpeng/exdata/graphics-devices.html

55 / 73

Showcasing56 / 73

First, globally set our theme!

You can set global theme with theme_set() and theme_update():

theme_set() takes a custom theme as a argument (e.g., theme_bw())

theme_update() takes individual theme elements as arguments and updates the current theme

theme_get() returns the current theme

theme_set(
  theme_classic(
    base_family = "Open Sans",
    base_size = 15
  )
)
theme_update(
  plot.margin = margin(.8, 1, .7, .8, "cm"),
  plot.title = element_text(
    family = "Roboto Slab",
    size = 20,
    margin = margin(b = .5, unit = "cm")
  ),
  plot.title.position = "plot",
  axis.title.x = element_text(
    margin = margin(t = .5, unit = "cm")
  )
)

57 / 73

Minimal Designs58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv")

  # A tibble: 4,516 x 10
     Value Item    Cond   Group Subject          Time Answer Accuracy Type   logRT
     <chr> <chr>   <chr>  <chr> <chr>           <dbl> <chr>     <dbl> <chr>  <dbl>
   1 Yes   Awaken~ Subje~ B     5cfa77e907f7dd~   917 No            0 Criti~  6.82
   2 No    Awaken~ Subje~ B     5c1d21196036e4~  2624 No            1 Criti~  7.87
   3 No    Awaken~ Subje~ B     5d730aab6b4b16~  2809 No            1 Criti~  7.94
   4 No    Awaken~ Subje~ B     5ddef0fae454cd~  1587 No            1 Criti~  7.37
   5 No    Awaken~ Subje~ B     5e7a8e8b11bf56~  1697 No            1 Criti~  7.44
   6 No    Awaken~ Verb   A     5d4613a2da9cb6~  1484 No            1 Criti~  7.30
   7 No    Awaken~ Verb   A     579e1e2b275be6~  1556 No            1 Criti~  7.35
   8 No    Awaken~ Verb   A     5dfe69ed11d879~  1699 No            1 Criti~  7.44
   9 No    Awaken~ Verb   A     5b9442cecd3808~  2159 No            1 Criti~  7.68
  10 No    Awaken~ Verb   A     5e7f3328849917~  1132 No            1 Criti~  7.03
  # ... with 4,506 more rows
58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical")

  # A tibble: 1,466 x 10
     Value Item    Cond   Group Subject          Time Answer Accuracy Type   logRT
     <chr> <chr>   <chr>  <chr> <chr>           <dbl> <chr>     <dbl> <chr>  <dbl>
   1 Yes   Awaken~ Subje~ B     5cfa77e907f7dd~   917 No            0 Criti~  6.82
   2 No    Awaken~ Subje~ B     5c1d21196036e4~  2624 No            1 Criti~  7.87
   3 No    Awaken~ Subje~ B     5d730aab6b4b16~  2809 No            1 Criti~  7.94
   4 No    Awaken~ Subje~ B     5ddef0fae454cd~  1587 No            1 Criti~  7.37
   5 No    Awaken~ Subje~ B     5e7a8e8b11bf56~  1697 No            1 Criti~  7.44
   6 No    Awaken~ Verb   A     5d4613a2da9cb6~  1484 No            1 Criti~  7.30
   7 No    Awaken~ Verb   A     579e1e2b275be6~  1556 No            1 Criti~  7.35
   8 No    Awaken~ Verb   A     5dfe69ed11d879~  1699 No            1 Criti~  7.44
   9 No    Awaken~ Verb   A     5b9442cecd3808~  2159 No            1 Criti~  7.68
  10 No    Awaken~ Verb   A     5e7f3328849917~  1132 No            1 Criti~  7.03
  # ... with 1,456 more rows
58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group)

  # A tibble: 1,466 x 10
  # Groups:   Cond, Group [4]
     Value Item    Cond   Group Subject          Time Answer Accuracy Type   logRT
     <chr> <chr>   <chr>  <chr> <chr>           <dbl> <chr>     <dbl> <chr>  <dbl>
   1 Yes   Awaken~ Subje~ B     5cfa77e907f7dd~   917 No            0 Criti~  6.82
   2 No    Awaken~ Subje~ B     5c1d21196036e4~  2624 No            1 Criti~  7.87
   3 No    Awaken~ Subje~ B     5d730aab6b4b16~  2809 No            1 Criti~  7.94
   4 No    Awaken~ Subje~ B     5ddef0fae454cd~  1587 No            1 Criti~  7.37
   5 No    Awaken~ Subje~ B     5e7a8e8b11bf56~  1697 No            1 Criti~  7.44
   6 No    Awaken~ Verb   A     5d4613a2da9cb6~  1484 No            1 Criti~  7.30
   7 No    Awaken~ Verb   A     579e1e2b275be6~  1556 No            1 Criti~  7.35
   8 No    Awaken~ Verb   A     5dfe69ed11d879~  1699 No            1 Criti~  7.44
   9 No    Awaken~ Verb   A     5b9442cecd3808~  2159 No            1 Criti~  7.68
  10 No    Awaken~ Verb   A     5e7f3328849917~  1132 No            1 Criti~  7.03
  # ... with 1,456 more rows
58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop')

  # A tibble: 4 x 3
    Cond    Group Accuracy
  * <chr>   <chr>    <dbl>
  1 Subject A        0.875
  2 Subject B        0.780
  3 Verb    A        0.846
  4 Verb    B        0.710
58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group))

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2)

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70"))

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70")) +
  labs(
    title = "Accuracy on Comprehension Task",
    x = "Pitch Accent Condition", y = NULL,
    fill = "Experiment Group"
  )

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70")) +
  labs(
    title = "Accuracy on Comprehension Task",
    x = "Pitch Accent Condition", y = NULL,
    fill = "Experiment Group"
  ) +
  coord_cartesian(ylim = c(0.5, 1))

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70")) +
  labs(
    title = "Accuracy on Comprehension Task",
    x = "Pitch Accent Condition", y = NULL,
    fill = "Experiment Group"
  ) +
  coord_cartesian(ylim = c(0.5, 1)) +
  guides(fill = guide_legend(direction = "horizontal", title.position = "top"))

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70")) +
  labs(
    title = "Accuracy on Comprehension Task",
    x = "Pitch Accent Condition", y = NULL,
    fill = "Experiment Group"
  ) +
  coord_cartesian(ylim = c(0.5, 1)) +
  guides(fill = guide_legend(direction = "horizontal", title.position = "top")) +
  theme(
    axis.ticks.x = element_blank(),
    axis.text.x = element_text(color = "black", margin = margin(t = .2, unit = "cm")),
    legend.position = c(.3, .93),
    plot.title = element_text(
      margin = margin(b = 1, unit = "cm")
    )
  )

58 / 73

read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70")) +
  labs(
    title = "Accuracy on Comprehension Task",
    x = "Pitch Accent Condition", y = NULL,
    fill = "Experiment Group"
  ) +
  coord_cartesian(ylim = c(0.5, 1)) +
  guides(fill = guide_legend(direction = "horizontal", title.position = "top")) +
  theme(
    axis.ticks.x = element_blank(),
    axis.text.x = element_text(color = "black", margin = margin(t = .2, unit = "cm")),
    legend.position = c(.3, .93),
    plot.title = element_text(
      margin = margin(b = 1, unit = "cm")
    )
  ) +
  scale_y_continuous(
    expand = expansion(0, 0),
    labels = percent_format(accuracy = 1)
  )

58 / 73

p <- read_csv("https://raw.githubusercontent.com/yjunechoe/Semantic-Persistence/master/processed.csv") %>%
  filter(Type == "Critical") %>%
  group_by(Cond, Group) %>%
  summarize(Accuracy = mean(Accuracy, na.rm = TRUE), .groups = 'drop') %>%
  ggplot(aes(x = Cond, y = Accuracy, fill = Group)) +
  geom_col(position = "dodge", color = "white", width = .7, size = 2) +
  scale_fill_manual(values = c("grey30", "grey70")) +
  labs(
    title = "Comprehension Task",
    x = "Pitch Accent Condition", y = NULL,
    fill = "Experiment Group"
  ) +
  coord_cartesian(ylim = c(0.5, 1)) +
  guides(fill = guide_legend(direction = "horizontal", title.position = "top")) +
  theme(
    axis.ticks.x = element_blank(),
    axis.text.x = element_text(color = "black", margin = margin(t = .2, unit = "cm")),
    legend.position = c(.3, .93),
    plot.title = element_text(
      margin = margin(b = 1.5, unit = "cm"),
      hjust = .5
    ),
    plot.margin = margin(.8, 1, .7, 1.2, "cm")
  ) +
  scale_y_continuous(
    expand = expansion(0, 0),
    labels = percent_format(accuracy = 1)
  )
library(cowplot)
ggdraw(p) +
  draw_label("Accuracy", x = .12, y = .88, fontfamily = "Open Sans", size = 14, color = "#2D2D2D")

58 / 73

Source: Husband & Patson (2020)

59 / 73

df <- crossing(level_1 = fct_inorder(c("Within", "Between")),
               level_2 = fct_inorder(c("Some", "Number", "Or")),
               level_3 = factor(c("Strong", "Weak")))
df$barheight <- c(.63, .35, .72, .55, .61, .15, .60, .55, .52, .63, .17, .16)
df %>% 
  ggplot(aes(level_3, barheight)) +
  geom_col(
    aes(fill = level_3),
    show.legend = FALSE
  ) +
  geom_errorbar(
    aes(ymin = barheight - .05, ymax = barheight + .05),
    width = .1) +
  facet_grid(level_2 ~ level_1) + #<<<
  theme_bw() +
  scale_fill_manual(values = c('grey40', 'grey80')) +
  ylim(0, 1) +
  labs(
    y = "Proportion of Strong Responses",
    x = "Prime Type") +
  theme_bw()

60 / 73

(Code)

61 / 73

(Code)

62 / 73

Plot composition63 / 73

# 3 column layout with `wrap_plots()`
plot_list <- list(state_election_plot, state_election_plot_A, state_election_plot_B,
                  state_election_plot_C, state_election_plot_D1, state_election_plot_D2)
patchwork::wrap_plots(plot_list, ncol = 3)

64 / 73

# You can also use operator symbols for layout
(plot_list[[1]] + plot_list[[2]])/plot_list[[3]]

65 / 73

# Or set up a custom layout with a grid
layout_design <- "ABDE
                  CCDE
                  CCDF"
wrap_plots(plot_list, design = layout_design)

66 / 73

Plot

# A more complete example with themes and annotations
plotter_fun <- function(x) {map2(1:3, rep(list(list(rnorm(100), rnorm(100))), 3), ~ {
  ggplot(NULL, aes(x = .y[[1]], y = .y[[2]])) +
    list(geom_point(), geom_density2d(), geom_density_2d_filled(show.legend = FALSE))[[.x]] +
    labs(x = NULL, y = NULL) +
    coord_cartesian(expand = FALSE) +
    theme_minimal()
})}
wrap_plots(c(plotter_fun(), plotter_fun()), ncol = 3) &
  theme(
    plot.margin = margin(.9, .2, .2, .2, unit = 'cm'),
    plot.tag = element_text(family = "Roboto Mono", size = 12, face = 'bold'),
    plot.tag.position = c(.1, 1.12),
    axis.text = element_text(color = 'black')
  ) &
  plot_annotation(
    title = "Plot Composition",
    tag_levels = list(paste0("Fig.", rep(1:2, each = 3), letters[rep(1:3, 2)])),
    theme = theme(
      plot.title = element_text(size = 16, family = "Roboto Slab", face = 'bold',
                                hjust = 0.5, margin = margin(b = .1, unit = 'cm')),
      plot.margin = margin(.3, .3, .3, .3, unit = 'cm')
    )
  )

67 / 73

Animations68 / 73

read_csv("https://raw.githubusercontent.com/bodowinter/canadian_vowel_shift_analysis/master/processed_data/production_processed.csv") %>% 
  group_by(Gender, Vowel, BirthDecade = 10 * BirthYear %/% 10) %>% 
  summarize(across(F1:F2, mean), .groups = 'drop') %>% 
  filter(Gender == "F") %>% 
  ggplot(aes(F2, F1)) +
  geom_text(aes(1450, 615, label = as.character(BirthDecade)),
            color = "gray80", size = 48) +
  stat_chull(fill = NA, color = "black", linetype = 2) +
  geom_label(aes(label = Vowel), size = 6, family = "Charis SIL") +
  scale_x_reverse(position = "top") +
  scale_y_reverse(position = "right") +
  theme(
    plot.margin = margin(.5,.8, 1.2, 1, unit = "cm"),
    axis.title.x.top = element_text(margin = margin(b = .2, unit = "cm")),
    axis.title.y.right = element_text(angle = 0, vjust = 0.5, margin = margin(l = .3, unit = "cm"))
  ) +
  coord_cartesian(clip = 'off') + 
  labs(title = "Canadian Vowel Shift", caption = "Source: Kettig & Winter (2017)") +
  # animation components from {gganimate}
  transition_states(BirthDecade) +
  shadow_mark(
    alpha = .1,
    color = "grey",
    exclude_layer = c(1, 2)
  )

69 / 73

de Carvalho et al. (2017)

70 / 73

(Code)

71 / 73

Interactivity

72 / 73

Resources

light

heavy

73 / 73

Help

Keyboard shortcuts

↑, ←, Pg Up, k

Go to previous slide

↓, →, Pg Dn, Space, j

Go to next slide

Home

Go to first slide

End

Go to last slide

Number + Return

Go to specific slide

b / m / f

Toggle blackout / mirrored / fullscreen mode

Clone slideshow

Toggle presenter mode

Restart the presentation timer

?, h

Toggle this help