library(here) # manage file paths
library(socviz) # data and some useful functions
library(tidyverse) # your friend and mine
library(gapminder) # some data04 — How ggplot Thinks
January 31, 2024
What is tidy data?
Tidy data
What is tidy data?
Tidy data is in long format
Every column is a single variable
Grolemund & Wickham
Every row is a single observation
Grolemund & Wickham
Every cell is a single value
Grolemund & Wickham
But also for less good reasons
Spot the untidiness
But also for less good reasons
- 😠 More than one header row
- 😡 Mixed data types in some columns
- 💀 Color and typography used to encode variables and their values
Fix it before you import it
Prevention is better than cure!
An excellent article by Karl Broman and Kara Woo:
- Broman KW, Woo KH (2018) “Data Organization in Spreadsheets”.” The American Statistician 78:2–10
Data organization in spreadsheets
What we need our code to make
- Data represented by visual elements;
- like position, length, color, and size;
- Each measured on some scale;
- Each scale with a labeled guide;
- With the plot itself also titled and labeled.
Here’s the whole thing, start to finish
Flow of action
We’ll go through it step by step
Flow of action
ggplot’s flow of action
What we start with
ggplot’s flow of action
Where we’re going
ggplot’s flow of action
Core steps
ggplot’s flow of action
Optional steps
ggplot’s flow of action: required
Tidy data
ggplot’s flow of action: required
Aesthetic mappings
ggplot’s flow of action: required
Geom
p has data and mappings but no geom
p
This empty plot has no geoms.
Add a geom
p + geom_point() 
A scatterplot of Life Expectancy vs GDP
Try a different geom
p + geom_smooth() 
A scatterplot of Life Expectancy vs GDP
Build your plots layer by layer
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_smooth()
Life Expectancy vs GDP, using a smoother.
This process is additive
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point() + geom_smooth()
Life Expectancy vs GDP, using a smoother.
This process is additive
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_smooth()
This process is additive
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_smooth() +
geom_point()
Every geom is a function
Functions take arguments
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point() +
geom_smooth(method = "lm") 
Keep Layering
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point()
Keep Layering
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point() +
geom_smooth(method = "lm")
Keep Layering
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point() +
geom_smooth(method = "lm") +
scale_x_log10()
Fix the labels
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point()
Fix the labels
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point() +
geom_smooth(method = "lm")
Fix the labels
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y=lifeExp))
p + geom_point() +
geom_smooth(method = "lm") +
scale_x_log10(labels = scales::label_dollar())
Add labels, title, and caption
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point() +
geom_smooth(method = "lm") +
scale_x_log10(labels = scales::label_dollar()) +
labs(x = "GDP Per Capita",
y = "Life Expectancy in Years",
title = "Economic Growth and Life Expectancy",
subtitle = "Data points are country-years",
caption = "Source: Gapminder.")
What has gone wrong here?
p_out
Try again
p_out
Geoms can take many arguments
p_out
alpha for overplotting
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point(alpha = 0.3) +
geom_smooth(method = "lm") +
scale_x_log10(labels = scales::label_dollar()) +
labs(x = "GDP Per Capita",
y = "Life Expectancy in Years",
title = "Economic Growth and Life Expectancy",
subtitle = "Data points are country-years",
caption = "Source: Gapminder.")
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp,
color = continent,
fill = continent))
p + geom_point()
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp,
color = continent,
fill = continent))
p + geom_point() +
geom_smooth(method = "loess")
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp,
color = continent,
fill = continent))
p + geom_point() +
geom_smooth(method = "loess") +
scale_x_log10(labels = scales::label_dollar())
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point(mapping = aes(color = continent))
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point(mapping = aes(color = continent)) +
geom_smooth(method = "loess")
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point(mapping = aes(color = continent)) +
geom_smooth(method = "loess") +
scale_x_log10(labels = scales::label_dollar())
Geoms can take their own mappings
p <- ggplot(data = gapminder,
mapping = aes(x = gdpPercap,
y = lifeExp))
p + geom_point(mapping = aes(color = continent)) +
geom_smooth(method = "loess") +
scale_x_log10(labels = scales::label_dollar())
Guides and scales reflect aes() mappings
mapping = aes(color = continent, fill = continent)
Guides and scales reflect aes() mappings
mapping = aes(color = continent, fill = continent)
mapping = aes(color = continent)
Try to make a lineplot
p <- ggplot(data = gapminder,
mapping = aes(x = year,
y = gdpPercap)) +
geom_line()
p
Try to make a lineplot
p <- ggplot(data = gapminder,
mapping = aes(x = year,
y = gdpPercap)) +
geom_line(mapping = aes(group = country))
p
Try to make a lineplot
p <- ggplot(data = gapminder,
mapping = aes(x = year,
y = gdpPercap)) +
geom_line(mapping = aes(group = country))
p
Facet the plot
gapminder |>
ggplot(mapping =
aes(x = year,
y = gdpPercap))
Facet the plot
gapminder |>
ggplot(mapping =
aes(x = year,
y = gdpPercap)) +
geom_line(mapping = aes(group = country))
Facet the plot
gapminder |>
ggplot(mapping =
aes(x = year,
y = gdpPercap)) +
geom_line(mapping = aes(group = country)) +
facet_wrap(~ continent)
stat_ functions behind the scenes
p <- ggplot(data = midwest,
mapping = aes(x = area))
p + geom_histogram()`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Here the default stat_ function for this geom has to make a choice. It is letting us know we might want to override it.
stat_ functions behind the scenes
p <- ggplot(data = midwest,
mapping = aes(x = area))
p + geom_histogram(bins = 10)
- We can choose either the number of bins or the
binwidth
Compare two distributions
## Two state codes
oh_wi <- c("OH", "WI")
midwest |>
filter(state %in% oh_wi) |>
ggplot(mapping = aes(x = percollege,
fill = state)) +
geom_histogram(alpha = 0.5,
position = "identity")
- Here we do the whole thing in a pipeline using the pipe and the
dplyrverbfilter()to subset rows of the data by some condition. - Experiment with leaving the
positionargument out, or changing it to"dodge".
geom_density()
p <- ggplot(data = midwest,
mapping = aes(x = area))
p + geom_density()
geom_density()
p <- ggplot(data = midwest,
mapping = aes(x = area,
fill = state,
color = state))
p + geom_density(alpha = 0.3)
geom_density()
midwest |>
filter(state %in% oh_wi) |>
ggplot(mapping = aes(x = area,
fill = state,
color = state)) +
geom_density(mapping = aes(y = after_stat(ndensity)),
alpha = 0.4)
ndensityhere is not in our data! It’s computed. Histogram and density geoms have default statistics, but you can ask them to do more. Theafter_statfunctions can do this work for us.
geom_bar() wants to count up
p <- ggplot(data = titanic,
mapping = aes(x = fate,
y = percent,
fill = sex))
p + geom_bar(stat = "identity") 
- By default
geom_bar()tries to count up data by category. (Really it’s thestat_count()function that does this behind the scenes.) By sayingstat="identity"we explicitly tell it not to do that. This also allows us to use aymapping. Normally this would be the result of the counting up.
geom_bar() stacks bars by default
p <- ggplot(data = titanic,
mapping = aes(x = fate,
y = percent,
fill = sex))
p + geom_bar(stat = "identity",
position = "dodge") 
- Position arguments adjust whether the things drawn are placed on top of one another (
"stack"), side-by-side ("dodge"), or taken as-is ("identity").
A quick theme() adjustment
p <- ggplot(data = titanic,
mapping = aes(x = fate,
y = percent,
fill = sex))
p + geom_bar(stat = "identity",
position = "dodge") +
theme(legend.position = "top") 
- The
theme()function controls the styling of parts of the plot that don’t belong to its “grammatical” structure. That is, that are not contributing to directly representing data.
For convenience, use geom_col()
p <- ggplot(data = titanic,
mapping = aes(x = fate,
y = percent,
fill = sex))
p + geom_col(position = "dodge") +
theme(legend.position = "top")
geom_col()assumesstat = "identity"by default. It’s for when you want to directly plot a table of values, rather than create a bar chart by summing over one varible categorized by another.
Using geom_col() for thresholds
