Visualization with ggplot2

January 24, 2017

ggplot2

Yardstick for plotting data in R
tidyverse package
Virtual no limits to plot types
Add-on packages available
Plotting resources

Plot types

What we will learn

Requests to learn another plot type?

Plot types

1 discrete variable

1 continuous variable

Plot types

2 continuous variables

1 continuous variable + date variable

Let’s make sure you’re able to plot your data

library(tidyverse)
mpg %>% ggplot() + geom_point(aes(displ, hwy))

Let’s make sure you’re able to plot your data

If you see a plot, you’re ready to go!
If you do not, reinstall tidyverse and re-run the test code

install.packages('tidyverse')
library(tidyverse)
mpg %>% ggplot() + geom_point(aes(displ, hwy))

If it still didn’t work, install ggplot2 and re-run the test code

install.packages('ggplot2')
library(ggplot2)
mpg %>% ggplot() + geom_point(aes(displ, hwy))

And, read your data…the easy way

library(tidyverse)
donor <- read.csv('https://goo.gl/tm9JQ5')
police <- read.csv('https://goo.gl/nNAuDy')

How to plot data

Pick your data
Pick your chart type (Geom)
Program stuff like chart titles, axis names, axis types, and legends (Scales)
Make it pretty (Themes)
Keep tuning and reuse code

ggplot basics

Geoms

Every plot needs a ggplot() function and a geom layer
Separate layers and other plotting functions with +

ggplot() + geom_bar() # create bar and stacked bar plots
ggplot() + geom_histogram() # create histograms
ggplot() + geom_point() # create scatter plots
ggplot() + geom_line() # create line plots

Your data is an argument in the ggplot() function

donor %>% ggplot() + geom_bar()
ggplot(donor) + geom_bar()

Geoms

Map your data to the plot with aes()
- aes() can be an argument in ggplot() or the geom
- Many arguments in the aes() function
- Not necessary to name x and y in aes()

aes(x = NULL, y = NULL, color = NULL
  , fill = NULL, alpha = NULL, label = NULL
  , shape = NULL, size = NULL, group = NULL
  , linetype = NULL
  )

donor %>% ggplot() + geom_bar(aes(primary_general))
donor %>% ggplot(aes(primary_general)) + geom_bar()

Geoms

donor %>% ggplot() + geom_bar(aes(primary_general))
donor %>% ggplot(aes(primary_general)) + geom_bar()

Geoms

donor %>% ggplot(aes(primary_general, fill = primary_general)) + geom_bar()

Exercise - 7 minutes

Build a bar chart with police data that shows how many incidences occurred in district_sector
Only include these district_sector values: c('B', 'E', 'D', 'R', 'O', 'C', 'K')
Color the bars using event_clearance_group
Only include these event_clearance_group values: c('TRAFFIC RELATED CALLS', 'FRAUD CALLS', 'BURGLARY', 'BIKE')

Hint

# Use to include/exclude values
filter()

Exercise - 7 minutes

Build a bar chart with police data that shows how many incidences occurred in district_sector
Only include these district_sector values: c('B', 'E', 'D', 'R', 'O', 'C', 'K')
Color the bars using event_clearance_group
Only include these event_clearance_group values: c('TRAFFIC RELATED CALLS', 'FRAUD CALLS', 'BURGLARY', 'BIKE')

Hint

police %>%
  filter() %>%
  ggplot()

Exercise - 7 minutes

police %>% 
  filter(
    district_sector %in% c('B', 'E', 'D', 'R', 'O', 'C', 'K') &
    event_clearance_group %in% c('TRAFFIC RELATED CALLS', 'FRAUD CALLS', 'BURGLARY', 'BIKE')
    ) %>%
  ggplot(aes(district_sector, fill = event_clearance_group)) + geom_bar()

Scales

Axis, legend, and chart titles

Axis names

Keep names short-ish
Consider including units in axis name
It is possible you don’t need an axis name

2 options

# Option 1
labs(x = 'X Axis Title', y = 'Y Axis Title')

# Option 2
xlab('X Axis Title')
ylab('Y Axis Title')

Axis, legend, and chart titles

donor %>% 
  ggplot(aes(primary_general, fill = primary_general)) + 
  geom_bar() + 
  labs(x = 'Election Type', y = 'Donations (#)')

Axis, legend, and chart titles

Try adding axis names to the district_sector plot you made during the exercise

police %>% 
  filter(
    district_sector %in% c('B', 'E', 'D', 'R', 'O', 'C', 'K') &
    event_clearance_group %in% c('TRAFFIC RELATED CALLS', 'FRAUD CALLS', 'BURGLARY', 'BIKE')
    ) %>%
  ggplot(aes(district_sector, fill = event_clearance_group)) + geom_bar() + 
  labs(x = 'District Sectors', y = 'Incidences by Event\nClearance Group (#)')

Axis, legend, and chart titles

Legend names

Keep names short
Be mindful of what the legend is for
- Color? Fill? Size? etc.
Consider hiding the legend title for a clean look

labs(fill = '')
labs(fill = element_blank())
labs(fill = NULL)
labs(colour = 'Check out these colors')

Axis, legend, and chart titles

police %>% 
  filter(
    district_sector %in% c('B', 'E', 'D', 'R', 'O', 'C', 'K') &
    event_clearance_group %in% c('TRAFFIC RELATED CALLS', 'FRAUD CALLS', 'BURGLARY', 'BIKE')
    ) %>%
  ggplot(aes(district_sector, fill = event_clearance_group)) + geom_bar() + 
  labs(x = 'Election Type', y = 'Donations (#)', fill = element_blank())

Axis, legend, and chart titles

Try removing the legend name from your district_sector plot

police %>% 
  filter(district_sector %in% c('B', 'E', 'D', 'R', 'O', 'C', 'K')) %>%
  ggplot(aes(district_sector, fill = event_clearance_group)) + geom_bar() + 
  labs(x = 'District Sectors', y = 'Incidences by Event\nClearance Group (#)'
       , fill = element_blank())

Axis, legend, and chart titles

Try removing the legend name from your district_sector plot

Axis, legend, and chart titles

Chart titles and subtitles

Consider making titles (or subtitles) descriptive
Don’t over use subtitles

# Option 1
labs(title = NULL, subtitle = NULL)

# Option 2
ggtitle(title = NULL, subtitle = NULL)

Axis, legend, and chart titles

donor %>% 
  ggplot(aes(primary_general, fill = primary_general)) + 
  geom_bar() + 
  labs(
    x = 'Election Type'
    , y = 'Donations (#)'
    , fill = element_blank()
    , title = 'Number of Donations by Election Type'
    , subtitle = 'Over 6,000 full election cycle donations'
    )

Axis, legend, and chart titles

Try adding a title to the district_sector plot

Axis units

Change axis units
This is a good idea and you should do it when its relevant
Good for x and y axes

install.packages('scales')
library(scales)

scale_y_continuous(labels = percent) # Add a percentage sign to numbers on axis
scale_y_continuous(labels = dollar) # Add a dollar sign to numbers on axis
scale_y_continuous(labels = comma) # Add a comma to numbers on axis

Axis units

donor %>% 
  ggplot(aes(primary_general, fill = primary_general)) + 
  geom_bar() + 
  labs(
    x = 'Election Type'
    , y = 'Donations (#)'
    , fill = element_blank()
    , title = 'Number of Donations by Election Type'
    , subtitle = 'Over 6,000 full election cycle donations'
    ) + 
  scale_y_continuous(labels = comma)

Axis units

Themes

Change layer and background colors
Change fonts
Change plot borders/boundaries and ticks
Pre-built themes vs. custom themes
- ggplot2 themes
- ggthemes themes

install.packages('ggthemes')
library(ggthemes)

Pre-built themes

Selected ggplot2 themes

theme_classic()
theme_minimal()
theme_dark()

Selected ggthemes themes

theme_stata() + scale_colour_stata() # scale_fill_stata()
theme_economist() + scale_colour_economist() # scale_fill_economist()
theme_fivethirtyeight() + scale_color_fivethirtyeight() # scale_fill_fivethirtyeight()
theme_wsj() + scale_colour_wsj() # scale_fill_wsj()
theme_pander() + scale_colour_pander() # scale_fill_pander()
theme_hc(bgcolor = "darkunica") + scale_colour_hc("darkunica") # scale_fill_hc("darkunica")

Pre-built themes

donor %>% 
  ggplot(aes(primary_general, fill = primary_general)) + 
  geom_bar() + 
  labs(
    x = 'Election Type'
    , y = 'Donations (#)'
    , fill = element_blank()
    , title = 'Number of Donations by Election Type'
    , subtitle = 'Over 6,000 full election cycle donations'
    ) + 
  scale_y_continuous(labels = comma) +
  theme_economist() + 
  scale_fill_economist()

Pre-built themes

Economist theme

Pre-built themes

Try adding a theme to the district_sector plot

Selected ggplot2 themes

theme_classic()
theme_minimal()
theme_dark()

Selected ggthemes themes

theme_stata() + scale_colour_stata() # scale_fill_stata()
theme_economist() + scale_colour_economist() # scale_fill_economist()
theme_fivethirtyeight() + scale_color_fivethirtyeight() # scale_fill_fivethirtyeight()
theme_wsj() + scale_colour_wsj() # scale_fill_wsj()
theme_pander() + scale_colour_pander() # scale_fill_pander()
theme_hc(bgcolor = "darkunica") + scale_colour_hc("darkunica") # scale_fill_hc("darkunica")

Pre-built themes

Try adding a theme to the district_sector plot

Custom themes

You can create custom themes with theme()
Greater control over chart aesthestics
Use custom themes if pre-built themes don’t cut it

Custom themes

t <- theme(
  plot.title = element_text(size=14, face="bold", vjust=1)
  , plot.background = element_blank()
  , panel.grid.major = element_blank()
  , panel.grid.minor = element_blank()
  , panel.border = element_blank()
  , panel.background = element_blank()
  , axis.ticks = element_blank()
  , axis.text = element_text(colour="black", size=12)
  , axis.text.x = element_text(angle=45, hjust=1)
  , legend.title = element_blank()
  , legend.position = "none"
  , legend.text = element_text(size=12)
  )

Histograms, scatterplots, and line plots

Histograms: 1 continuous variable

donor %>% 
  filter(amount < 1000) %>%
  ggplot(aes(amount)) + geom_histogram()

Histograms: 1 continuous variable

donor %>% 
  filter(amount < 1000) %>%
  ggplot(aes(amount)) + geom_histogram()

geom_density() # works like geom_histogram(); continuous curve

Scatterplots: 2 continuous variables

donor %>% 
  ggplot(aes(receipt_year, election_year)) + geom_point()

Scatterplots: 2 continuous variables

donor %>% 
  ggplot(aes(receipt_year, election_year)) + geom_point()

geom_count() # adds a size component to your scatterplot
geom_jitter() # adjusts cartesian location of point relative to other points

Line plot: 1 continuous variable + date variable

donor %>% 
  ggplot(aes(receipt_year, amount)) + geom_line()

Exercise - 5 minutes

With either dataset, donor or police, create a scatterplot
Include two aes() arguments in addition to x and y
- For example, size, shape, or alpha
Start by identifying numeric variables in your dataset

Exercise - 5 minutes

police %>% 
  ggplot(aes(latitude, longitude, color = district_sector, alpha = district_sector)) + 
  geom_point() + 
  theme_classic()

Aggregate data for chart creation with group_by() and summarise()

What you can do with group_by() and summarise()

Count, sum, average, and identify max and min values
This is like COUNTIFS and SUMIFS formulas in Excel and LOD expressions in Tableau
Example questions you could answer
- In donor, which type received the most money in donations from 'REPUBLICAN'?
- In donor, which type received the most money in donations in the receipt_year 2015?
- In donor, what is the largest average donation amount for contributor_state?

How you use group_by() and summarise() to aggregate data

In group_by(), list the variables by which you want to aggregate data
In summarise(), create a variable and define the variable with an aggregation function
Use %>% to ‘link’ group_by() and summarise()
Aggregation functions
- n(), n_distinct(), sum(), mean(), max(), min(), etc.

Example
In donor, which type received the most money in donations from 'REPUBLICAN'?

donor %>% group_by(type, party) %>% summarise(dollars = sum(amount, na.rm = TRUE))

How you use group_by() and summarise() to aggregate data

Example
In donor, which type received the most money in donations from 'REPUBLICAN'?

donor %>% group_by(type, party) %>% summarise(dollars = sum(amount, na.rm = TRUE))
## # A tibble: 7 x 3
## # Groups: type [?]
##   type                party        dollars
##   <fctr>              <fctr>         <dbl>
## 1 Candidate           DEMOCRAT      311690
## 2 Candidate           INDEPENDENT      467
## 3 Candidate           NON PARTISAN  206196
## 4 Candidate           NONE            6465
## 5 Candidate           OTHER          25859
## 6 Candidate           REPUBLICAN    313540
## 7 Political Committee <NA>         1563098

How you use group_by() and summarise() to aggregate data

How would you turn the tabular output data into a chart?

donor %>% group_by(type, party) %>% summarise(dollars = sum(amount, na.rm = TRUE))
## # A tibble: 7 x 3
## # Groups: type [?]
##   type                party        dollars
##   <fctr>              <fctr>         <dbl>
## 1 Candidate           DEMOCRAT      311690
## 2 Candidate           INDEPENDENT      467
## 3 Candidate           NON PARTISAN  206196
## 4 Candidate           NONE            6465
## 5 Candidate           OTHER          25859
## 6 Candidate           REPUBLICAN    313540
## 7 Political Committee <NA>         1563098

How you use group_by() and summarise() to aggregate data

How would you turn the tabular output data into a chart?

donor %>% 
  group_by(type, party) %>% 
  summarise(dollars = sum(amount, na.rm = TRUE)) %>%
  ggplot(aes(type, dollars, fill = party)) + geom_bar(stat = "identity") + 
  coord_flip() + 
  scale_y_continuous(labels = dollar) + theme_classic()

How you use group_by() and summarise() to aggregate data

How would you turn the tabular output data into a chart?

donor %>% 
  group_by(type, party) %>% 
  summarise(dollars = sum(amount, na.rm = TRUE)) %>%
  ggplot(aes(type, dollars, fill = party)) + geom_bar(stat = "identity") + 
  scale_y_continuous(labels = dollar) + theme_classic()

Necessary when data is aggregated

geom_bar(stat = 'identity') # use stat when working with aggregated data
geom_line(stat = 'identity')

How you use group_by() and summarise() to aggregate data

Example
In donor, which type received the most money in donations in the receipt_year 2015?

donor %>% 
  group_by(receipt_year, type) %>% 
  summarise(total_amount = sum(amount, na.rm = TRUE)) %>%
  filter(receipt_year %in% 2015)
## # A tibble: 2 x 3
## # Groups: receipt_year [1]
##   receipt_year type                total_amount
##          <int> <fctr>                     <dbl>
## 1         2015 Candidate                  55729
## 2         2015 Political Committee        66821

How you use group_by() and summarise() to aggregate data

How would you turn the tabular output data into a chart?

## # A tibble: 5 x 3
##   receipt_year type                total_amount
##          <int> <fctr>                     <dbl>
## 1         2013 Political Committee       118804
## 2         2008 Candidate                 139377
## 3         2007 Political Committee       162470
## 4         2017 Political Committee        44280
## 5         2013 Candidate                  46316

How you use group_by() and summarise() to aggregate data

How would you turn the tabular output data into a chart?

donor %>% 
  group_by(receipt_year, type) %>% 
  summarise(total_amount = sum(amount, na.rm = TRUE)) %>%
  ggplot(aes(receipt_year, total_amount, color = type)) +
  geom_line(stat = 'identity') + 
  scale_y_continuous(labels = dollar) + theme_classic()

Exercise - 10 minutes

In donor, determine what the largest average donation amount is for contributor_state?
Use mean() in the summarise function
Exclude NA values with filter()
Show your findings in a chart

donor %>%
  group_by() %>%
  summarise() %>%
  filter()

Exercise - 10 minutes

In donor, determine what the largest average donation amount is for contributor_state?
Use mean() in the summarise function
Exclude NA values with filter()
Show your findings in a chart

donor %>%
  group_by() %>%
  summarise() %>%
  filter() %>%
  ggplot() %>%
  geom_bar()

Exercise - 10 minutes

In donor, determine what the largest average donation amount is for contributor_state?
Use mean() in the summarise function
Exclude NA values with filter()
Show your findings in a chart

donor %>%
  group_by() %>%
  summarise() %>%
  filter() %>%
  ggplot() %>%
  geom_bar(stat = 'identity')

Exercise - 10 minutes

donor %>% 
  group_by(contributor_state) %>% 
  summarise(avg_amount = mean(amount, na.rm = TRUE)) %>%
  filter(! contributor_state %in% NA) %>%
  ggplot(aes(reorder(contributor_state, -avg_amount), avg_amount)) +
  geom_bar(stat = 'identity') + 
  scale_y_continuous(labels = dollar) + theme_classic()

Exercise - 15 minutes

Create a chart from the police dataset
Make sure your axes are labeled correctly
Give your chart a title
Use color in aes()
Use a pre-built theme

Exercise - 15 minutes

police %>% 
  filter(
    ! event_clearance_ampm %in% NA & 
    ! event_clearance_group %in% NA
    ) %>%
  group_by(event_clearance_group, event_clearance_ampm) %>%
  summarise(n = n()) %>%
  ungroup %>%
  group_by(event_clearance_group) %>%
  mutate(total = sum(n, na.rm = TRUE)) %>%
  filter(total >= 100) %>%
  ggplot(aes(reorder(event_clearance_group, n), n, fill = event_clearance_ampm)) + 
  geom_bar(stat = 'identity') + 
  coord_flip() + 
  theme_wsj() + 
  scale_colour_wsj() + 
  labs(title = 'Most incidence occur\nin the AM', fill = 'Time of Day')

Exercise - 15 minutes

police %>% 
  filter(
    ! event_clearance_ampm %in% NA & 
    ! event_clearance_group %in% NA
    ) %>%
  group_by(event_clearance_group, event_clearance_ampm) %>%
  summarise(n = n()) %>% 
  ggplot(aes(n, color = event_clearance_ampm)) + 
  geom_density(size = 1) + 
  theme_wsj() + 
  scale_colour_wsj() + 
  labs(title = 'Most incidences have\noccurred fewer than 25 times', colour = 'Time of Day')

Exercise - 10 minutes

Create a chart from the donor dataset
Make some comparison between values in cash_or_in_kind
Give your chart a title
Use a pre-built theme

Exercise - 10 minutes

donor %>% 
  ggplot(aes(cash_or_in_kind, fill = primary_general)) + 
  geom_bar() + 
  theme_wsj() + 
  scale_colour_wsj() + 
  labs(title = 'So many more\ncash donations!!!', colour = 'Election Type')