This practical will teach you the basics of ggplot2. It is split in 4 parts, each dedicated to a chart family:

Correlation: scatterplot and bubble plot
Distribution: histogram, density and boxplot.
Ranking: barplot, lollipop and treemap.
Evolution: line plot and area chart

Get ready

→ I strongly suggest to use R Studio, the best way to develop R code.

→ Start a new script with File → new file → R Script. Save this file somewhere in your computer. You will store all your command lines in it.

→ This practical requires several R packages.

Q0.1 Install the ggplot2 package if needed and load it. ggplot2 is a very powerful package for data visualization with R. It is the main topic of this practical.

# Install the package if needed
#install.packages("ggplot2")

# Load it
library(ggplot2)

Q0.2 Install and load dplyr. dplyr is part of the tidyverse and is very useful for data manipulation. It also provides the %>% operators that allows to ‘pipe’ commands.

# Install the package if needed
#install.packages("dplyr")

# Load it
library(dplyr)

1- Correlation

The first part of this practical will teach you how to build scatterplots and bubble charts: the 2 most common chart types to visualize correlations.

Q1.1 Load the gapminder dataset stored in the gapminder package. Have a look to the 6 first rows using the head() function. Describe briefly what you see as comments in your script. Check how many rows are available with nrow()

# Install the package if needed
#install.packages("gapminder")

# Load it
library(gapminder)

# Have a look to the first rows
head(gapminder)

# How many rows?
# nrow(gapminder)

Q1.2 How many years are available in this dataset? How many data-points for each year? Full code is provided for this question.

# Number of different year?
gapminder %>%
  select(year) %>%
  unique() %>%
  nrow()
# or
length(unique(gapminder$year))

# Number of country available per year?
gapminder %>%
  group_by(year) %>%
  summarize( n=n() )

Q1.3 Build a scatterplot showing the relationship between gdpPercap and lifeExp in 1952. Use geom_point(). What do you observe?

# basic scatterplot
gapminder %>%
  filter(year=="1952") %>%
  ggplot( aes(x=gdpPercap, y=lifeExp)) +
    geom_point()

Q1.4 On the previous chart, one country is very different. Which one is it?

You should get something like:

# Number of different year?
gapminder %>%
  filter(year=="1952" & gdpPercap>90000)

Q1.5 Build the same chart, but get rid of this country. What trend do you observe? Does it make sense? What’s missing? What could be better?

You should get a chart like this:

# basic scatterplot
gapminder %>%
  filter(year=="1952" & country!="Kuwait") %>%
  ggplot( aes(x=gdpPercap, y=lifeExp)) +
    geom_point()

Q1.6 Color dots according to their continent. In the aes() part of the code, use the color argument.

gapminder %>%
  filter(year=="1952" & country!="Kuwait") %>%
  ggplot( aes(x=gdpPercap, y=lifeExp, color=continent)) +
    geom_point()

Q1.7 Let’s observe an additional variable: make the circle size proportionnal to the population (pop). This is done with the size argument of aes(). How do you call this kind of chart?

gapminder %>%
  filter(year=="1952" & country!="Kuwait") %>%
  ggplot( aes(x=gdpPercap, y=lifeExp, color=continent, size=pop)) +
    geom_point()

Bonus You’re in advance? Try to do the following:

custom the general appearance using the theme_ipsum of the hrbrthemes library.
add transparency to circles to limit overlapping impact, with the alpha argument of aes
sort your data by pop size to put the small circle on top of the chart, not hidden by big bubbles
use the ggplotly() function of the plotly package to make this chart interactive

You should get something like this chart:

# Additionnal packages:
library(hrbrthemes) # for general style
library(plotly)     # to make the chart interactive

# Chart
p <- gapminder %>%
  filter(year=="1952" & country!="Kuwait") %>%
  arrange(desc(pop)) %>%
  ggplot( aes(x=gdpPercap, y=lifeExp, fill=continent, size=pop)) +
    geom_point(alpha=0.7, stroke="white", shape=21) +
    theme_ipsum()

# Interactive more
ggplotly(p)

2- Distribution

This second part is dedicated to the visualization of distribution. It is split in 2 parts:

Visualizing one distribution
Comparing distribution for several groups or variables

2.1 - One distribution

The example dataset provides the AirBnb night prices of ~1000 appartments on the French Riviera. Data is stored on Github and can be loaded in R as follow:

# Load dataset from github
data <- read.table("https://raw.githubusercontent.com/holtzy/data_to_viz/master/Example_dataset/1_OneNum.csv", header=TRUE)

Q2.1.1 How many rows in the dataset? (use nrow()) What is the min? The max? (use summary()). Do you see anything weird? What kind of chart would you build to visualize this kind of data?

# nrow(data)
# summary(data)

Q2.1.2 Build a histogram of the data with geom_histogram(). Are you happy with the output? How can we improve it?

data %>%
  ggplot( aes(x=price)) +
    geom_histogram()

Q2.1.3 Build a histogram without prices over 1500 euros. ggplot2 displays a warning message, why? What does it mean? What’s the main caveat of histograms?

data %>%
  filter(price<1500) %>%
  ggplot( aes(x=price)) +
    geom_histogram()

Q2.1.4 Build the histogram with different values of binwidth, for prices <400. What do you observe? Is it important to play with this parameter?

data %>%
  filter(price<400) %>%
  ggplot( aes(x=price)) +
    geom_histogram(binwidth = 2)

Q2.1.5 Use geom_density() to build a density chart. Use the fill argument to set the color. Use the help() function to find out what is the equivalent of bin_width for density chart? Check its effect using different values.

data %>%
  filter(price<1000) %>%
  ggplot( aes(x=price)) +
    geom_density(color="transparent", fill="#69b3a2", bw=5)

2.2 - Several distributions

Dataset: questions like What probability would you assign to the phrase Highly likely were asked. Answers were given in the range 0-100. It allows to understand how people perceive probability vocabulary. Data is stored on Github and can be loaded in R as follow:

# Load dataset from github
data <- read.table("https://raw.githubusercontent.com/holtzy/Teaching/master/DATA/probability.csv", header=TRUE, sep=",")

Q2.2.1 As usual, check data main features with nrow(), summary() or any other function you think is useful.

# Data size?
#nrow(data)

# occurence of each word:
#table(data$text)

Q2.2.2 What kind of chart would you do to compare the 8 categories?

Q2.2.3 Build a basic boxplot using the default options of geom_boxplot()

ggplot(data, aes(x=text, y=value, fill=text)) +
    geom_boxplot()

Q2.2.4 What do you observe? Can you improve this chart? What would you change? Do you remind what the different parts of the box mean?

Q2.2.5 Apply the following modifications to the previous boxplot:

order groups by increasing median of value. This is done thanks to the forcats package. Code is provided.
flip X and Y axis (coord_flip())
get rid of the legend (theme)

# Library forcats to reorder data
library(forcats)

# Reorder data
data %>%
  mutate(text = fct_reorder(text, value, .fun = median)) %>%
  ggplot(aes(x=text, y=value, fill=text)) +
    geom_boxplot() +
    theme(
      legend.position = "none"
    ) +
    coord_flip()

Q2.2.6 What is the main caveat with boxplot? How can we do better?

Q2.2.7 Let’s show individual data points using the geom_jitter() function. Explain what this function exactly does. Try to get a nice output using the width, size, alpha and color options.

# Library forcats to reorder data
library(forcats)

# Reorder data
data %>%
  mutate(text = fct_reorder(text, value, .fun = median)) %>%
  ggplot(aes(x=text, y=value, fill=text)) +
    geom_boxplot() +
    geom_jitter(color="grey", width=.4, size=.5, alpha=.8) +
    theme(
      legend.position = "none"
    ) +
    coord_flip()

Bonus Too fast? Try to do the following:

build a violin plot with geom_violin()
find out how to add a red circle to represent the mean of each group
search the internet to build a ridgeline chart.

3- Ranking

Let’s talk about the quantity of weapons exported by the top 50 largest exporters in 2017 (source). The dataset is available on github. Load it in R:

# Load dataset from github
data <- read.table("https://raw.githubusercontent.com/holtzy/data_to_viz/master/Example_dataset/7_OneCatOneNum.csv", header=TRUE, sep=",")

Q3.1 Have a quick look to the dataset. Describe it. What kind of chart can you build with this dataset? Which one would be the best in your opinion? What are the countries on top of the ranking?

#head(data)
#nrow(data)
#data %>% 
#  arrange(desc(Value)) %>%
#  head(5)

Q3.2 Start with a basic barplot using geom_bar().

Note: by default geom_bar() takes only one categorical variable as input, used for the x axis. It counts the number of cases at each x position and display it on the Y axis. In our case, we want to provide a y value for each group. This is why we need to specify stat="identity".

data %>%
  ggplot( aes(x=Country, y=Value) ) +
    geom_bar(stat="identity")

Q3.3 Color all bars with the same color: #69b3a2. Don’t like the color? Pick another one. Do you have to use fill or color? Why?

data %>%
  ggplot( aes(x=Country, y=Value) ) +
    geom_bar(stat="identity", fill="#69b3a2")

Q3.4 Set a different color for each bar. Do you like the output? Is it useful? Do you understand the difference between adding an option inside or outside aes() ?

data %>%
  ggplot( aes(x=Country, y=Value, fill=Country) ) +
    geom_bar(stat="identity")

Q3.5 Previous barplots are a bit disappointing aren’t they? What can you improve?

Q3.6 Try the following:

use coord_flip() to get a horizontal version
check the code below to reorder countries.

You should get that kind of output:

data %>%
  filter(!is.na(Value)) %>%
  arrange(Value) %>%
  mutate(Country=factor(Country, Country)) %>%
  ggplot( aes(x=Country, y=Value) ) +
    geom_bar(stat="identity", fill="#69b3a2") +
    coord_flip() +
    xlab("")

Q3.7 A lollipop plot is used in the same conditions as a barplot. Build it with:

geom_segment() for the stems. Arguments needed are x, xend, y and yend.
geom_point() for the circles. Needs x and y only.

You should get:

data %>%
  filter(!is.na(Value)) %>%
  arrange(Value) %>%
  mutate(Country=factor(Country, Country)) %>%
  ggplot( aes(x=Country, y=Value) ) +
    geom_segment( aes(x=Country ,xend=Country, y=0, yend=Value), color="grey") +
    geom_point(size=3, color="#69b3a2") +
    coord_flip() +
    xlab("")

BONUSTry the following:

Improve the previous lollipop chart with the theme_ipsum. Be creative to make it even better.
Visit the treemap section of the R graph gallery. Apply this kind of chart to the weapon dataset. You should get the chart below.
Ask google what a circular barchart is. What are the pros and cons of this chart compared to the classic barplot? Try to implement it with R and ggplot2.

# Package
library(treemap)

# Plot
treemap(data,
            
            # data
            index="Country",
            vSize="Value",
            type="index",
            
            # Main
            title="",
            palette="Dark2",

            # Borders:
            border.col=c("black"),             
            border.lwds=1,                         
        
            # Labels
            fontsize.labels=0.5,
            fontcolor.labels="white",
            fontface.labels=1,            
            bg.labels=c("transparent"),              
            align.labels=c("left", "top"),                                  
            overlap.labels=0.5,
            inflate.labels=T                        # If true, labels are bigger when rectangle is bigger.

            
            )

4- Evolution

Let’s consider the evolution of the bitcoin price between April 2013 and April 2018. Data are stored on github. Load the dataset using the following code:

# Load dataset from github
data <- read.table("https://raw.githubusercontent.com/holtzy/data_to_viz/master/Example_dataset/3_TwoNumOrdered.csv", header=T)
data$date <- as.Date(data$date)
#Here is one change

Q4.1 Build a basic line chart showing the bitcoin price evolution using geom_line().

data %>%
  ggplot( aes(x=date, y=value)) +
    geom_line(color="#69b3a2")

Q4.2 Switch to an area chart using geom_area(). Use the color and fill argument to customize chart colors.

data %>%
  ggplot( aes(x=date, y=value)) +
    geom_area(color="#69b3a2", fill="#69b3a2")

Q4.3 Select the last 10 values using tail(). Build a connected scatterplot using geom_point(), geom_line() and geom_area().

data %>%
  tail(10) %>%
  ggplot( aes(x=date, y=value)) +
    geom_area(fill="#69b3a2", alpha=0.5) +
    geom_line(color="#69b3a2") +
    geom_point()

Bonus Visit the time series section of the R graph gallery. Try to use the HTML widget called dygraph to build an interactive version of this lineplot.

You should get something like this.

# Library
library(dygraphs)
library(xts)          # To make the convertion data-frame / xts format
library(lubridate)

# Then you can create the xts format
don <- xts(x = data$value, order.by = data$date)

# graph
dygraph(don) %>%
  dyOptions(labelsUTC = TRUE, fillGraph=TRUE, fillAlpha=0.1, drawGrid = FALSE, colors="#D8AE5A") %>%
  dyRangeSelector() %>%
  dyCrosshair(direction = "vertical") %>%
  dyHighlight(highlightCircleSize = 5, highlightSeriesBackgroundAlpha = 0.2, hideOnMouseOut = FALSE)  %>%
  dyRoller(rollPeriod = 1)

A work by a practical by Yan Holtz

yan.holtz.data@gmail.com

Basic dataviz with R and ggplot2

a practical by Yan Holtz - 07 avril 2019