Introduction to ggplot2

Visualizing data

Why visualize data?

Why visualize data?

• Data visualization is not a decoration; it is an analytical instrument.

• Human cognition is optimized for pattern recognition in visual space, not for reading tables or numbers.

Why visualize data?

Why do we use R but don’t learn the basic first?

• Time limitations.
• You can still do it! I personally know 2 people that did it (small smaple size i know…).
• Encourage you to learn R

Steps of Visualization

1. Get to know your data
2. What is your message?
   • What do you want to show?
3. What types of variables do you have?
   • Numeric: continuous (i.e. price, weight, distance etc) , or discrete (i.e. age, size of a family etc)?
   • Non-numeric: nominal (i.e. gender, race, t-shirt color etc), ordinal (i.e. t-shirt size, schooling degree, satisfaction degree)
4. What kind of plot can you use?
5. Make the plot, understand it
6. Make it beautiful

Use the Chart Decision Table

Lets Write Our first program!

Hello World!

print("Hello world!")

[1] "Hello world!"

Can you introduce yourself to R?

print("Hello R!, I am Mücahit")

[1] "Hello R!, I am Mücahit"

Well Done!

Before going into the data

Necessary things

We first need to install the packages we want to use.

install.packages("tidyverse") 
install.packages("janitor")

1

Run these only if you never run them before. Installing a package once is enough.

Read the data

library(tidyverse)
library(janitor)
read.csv("https://raw.githubusercontent.com/mucahitzor/IKT2010/refs/heads/main/data/earnings.csv") %>%
  as_tibble() -> earnings

1

Use the tidyverse, and janitor, libraries so we can use the the functions that come with them

2

My data is a csv file in this path (a path is where the data is), read it, and then do the following ( %>% )

3

show it as a tibble: a tibble is a nice way to see our data. With -> earnings we are saying that save this data set into our memory so we can use it whenever we want.

Have a look at the data

lets see how does the data look like. There are two ways:

earnings

1

Just writing the name of the data shows us how it looks like on the console

# A tibble: 200 × 3
      id gender income
   <int> <chr>   <dbl>
 1     1 Male   19376.
 2     2 Male    9540.
 3     3 Male   10501.
 4     4 Male   26262.
 5     5 Male   23313.
 6     6 Male   15835.
 7     7 Male   12728.
 8     8 Female 22247.
 9     9 Female 41456.
10    10 Male   16254.
# ℹ 190 more rows

earnings %>% view()

1

take my dataset, and then make me view it visually in more detail

Understanding Our Data

What are our variables?

earnings

1

lets see how our data look like again

To see the variable names and information on numerical variables we use the function summary()

So we need to tell R to

• take my data set named earnings

• and do the thing i will tell you on the next line (%>%)

• give me the summary() of the data set.

earnings %>% 
  summary()

       id            gender              income     
 Min.   :  1.00   Length:200         Min.   : 9448  
 1st Qu.: 50.75   Class :character   1st Qu.:15920  
 Median :100.50   Mode  :character   Median :21052  
 Mean   :100.50                      Mean   :22892  
 3rd Qu.:150.25                      3rd Qu.:27321  
 Max.   :200.00                      Max.   :53647  

Understanding Our Data

What are our variables?

For non-numeric variables (like gender) we tell R to.

Take my data set named earnings and do the following ( %>% ).

apply tabyl({variable name}) to show statistics about the variable name we want to see. We want to apply tabyl() function our non-numeric variable named gender.

earnings %>%
  tabyl(gender)

1

Take my data set named earnings.

2

using the tabyl() function, show me statistics of variable gender.

 gender   n percent
 Female  91   0.455
   Male 109   0.545

Understanding Our Data

Our Goal Today: How is income distributed?

• We are trying to find out how income is distributed.

• Income is a continuous variable.

• To see how does a continuous variable is distributed: use a histogram or a density plot.

How is income distributed?

Distribution of one continuous variable

Histogram

Lets have a look at the data again

# A tibble: 200 × 3
      id gender income
   <int> <chr>   <dbl>
 1     1 Male   19376.
 2     2 Male    9540.
 3     3 Male   10501.
 4     4 Male   26262.
 5     5 Male   23313.
 6     6 Male   15835.
 7     7 Male   12728.
 8     8 Female 22247.
 9     9 Female 41456.
10    10 Male   16254.
# ℹ 190 more rows

• We see the first 10 individuals’ income values.
• Each row (observation) is one person.
• But we have 200 observations that we need to somehow understand how income is distributed.
• This is impossible just looking at the data in a table.
• We can have a look at the summary statistics to see the min, quartiles, median, mean, and max. But this still does not show us the shape.

earnings %>%
  summary()

       id            gender              income     
 Min.   :  1.00   Length:200         Min.   : 9448  
 1st Qu.: 50.75   Class :character   1st Qu.:15920  
 Median :100.50   Mode  :character   Median :21052  
 Mean   :100.50                      Mean   :22892  
 3rd Qu.:150.25                      3rd Qu.:27321  
 Max.   :200.00                      Max.   :53647  

How is income distributed?

Distribution of one continuous variable

Histogram

• A histogram groups continuous data into intervals (bins) and counts how many observations fall inside each interval.

• A bin is basically an interval. For example, each of the following is a bin with width of 1000.

  • “0-1000”
  • “1000-2000”
  • “2000-3000”

• We need to group the continuous data into bins and count how many of the observations belong to each bin.

• Choosing binwidth is somewhat depends on situation, try different binwidths and you’ll have different shapes.

How is income distributed?

Distribution of one continuous variable

Histogram

earnings %>% 
  summary()

       id            gender              income     
 Min.   :  1.00   Length:200         Min.   : 9448  
 1st Qu.: 50.75   Class :character   1st Qu.:15920  
 Median :100.50   Mode  :character   Median :21052  
 Mean   :100.50                      Mean   :22892  
 3rd Qu.:150.25                      3rd Qu.:27321  
 Max.   :200.00                      Max.   :53647  

• Lets start drawing a histogram of income by hand with 1000 bin width with our hand. Our bins will start from the minimum value of income.

  • “9000-10.000”, “10.000-11.0000”, “11.000-12.000”…

• For each observation we look at the income of the person and identify which bin it belongs to and start counting. As we see more observations falling into the same bin, the height of the bin (count) increases.

• Before doing this we first need to place our x and y axises.

  • Draw one x axis and one y axis,
  • On x axis we have income, in bins of 1000, starting from 9000.
  • On y axis we always have count, showing the number of observations that belongs to each bin.

How is income distributed?

Distribution of one continuous variable

Histogram

Lets plot the first 10 observations.

# A tibble: 200 × 3
      id gender income
   <int> <chr>   <dbl>
 1     1 Male   19376.
 2     2 Male    9540.
 3     3 Male   10501.
 4     4 Male   26262.
 5     5 Male   23313.
 6     6 Male   15835.
 7     7 Male   12728.
 8     8 Female 22247.
 9     9 Female 41456.
10    10 Male   16254.
# ℹ 190 more rows

• It is very tiring to go over all 200 observations.

• Instead we can tell R to do it for us!

How is income distributed?

Distribution of one continuous variable

Histogram

Take my data set named earnings,

earnings

I want to make a visualization, take my data set and apply ggplot.

What we will have is a blank piece of paper. We did not specified our axises yet.

earnings %>% 
  ggplot()

How is income distributed?

Distribution of one continuous variable

Histogram

Similar to drawing with hand, we first draw the x and y axis. To do this with R we use the function aes(x ={ variable name}). And we add this into our existing blank page using + symbol.

+ will always be at the end of the lines.

earnings %>%
  ggplot() +
  aes(x=income)

1

Take my data set named aernings

2

I want to make a vizualization

3

My variable on x axis is income

For histograms, we only need to give the name of the x variable, R will count each observations by itself!

How is income distributed?

Distribution of one continuous variable

Histogram

Now it is time to tell R to make a histogram.

We use geom_{plot type} to tell R what kind of plot we want.

For example, to make a histogram, we use geom_histogram().

To make a density plot, we use geom_density().

To make a bar plot, we use geom_bar().

To make a scatter point plot , we use geom_point().

How is income distributed?

Distribution of one continuous variable

Histogram

• Now lets tell R to
  • Use my data set named earnings and then ( %>% ),
  • Use ggplot() to make a vizualization, +,
  • My x variable is income, and on top of that i am adding (+),
  • A histogram.

How is income distributed?

Distribution of one continuous variable

Histogram

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram()

How is income distributed?

Distribution of one continuous variable

Histogram

How is income distributed?

Distribution of one continuous variable

Histogram

First fix this awful plot with start by adding a theme on top of our visualization. theme_{theme name} allows us to use a theme. Make sure theme is always at the last line.

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram() +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

By default R chooses some binwidth by itself inside geom_histogram(), even though we don’t see it geom_histogram(binwidth = 30). We want to change binwidth to what we want: binwidth=1000

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000) +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

R also chooses which color to fill the bins with. We want to change the color of the filling to lightskyblue inside geom_histogram(binwidth = 1000, fill={"name of the color"}).

Note that we use , to separate each argument.

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000, fill = "lightskyblue") +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

To change the color of the edges of the bins we use color={name of the color} argument inside geom_histogram(). Lets change it to black.

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000, fill = "lightskyblue", color = "black") +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

• Now, it is your time to:
  • Choose a different binwidth to see how the shape changes,
  • Choose another fill, and edge color,
  • Choose another theme
    • To see the theme options, delete the theme theme_classic() we set earlier and start typing theme_ R will show you suggestions.

How is income distributed?

Distribution of one continuous variable

Histogram

• Before we move on to interpret the plot, i want to increase the number of values in the x axis. Currently we have 5 values which are not really enough to see the distribution.

• To increase the number of values from 5 to 10 on the x axis, use scale_x_continuous(n.breaks=10) layer.

• Remember to add this before the theme!

How is income distributed?

Distribution of one continuous variable

Histogram

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000, fill = "lightskyblue", color = "black") +
  scale_x_continuous(n.breaks=10) +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

We have the same problem in y axis as well.

How can we increase the number of values in y axis? For x axis the function was scale_x_continuous(n.breaks=10). Can you set the number of breaks on y axis to 10?

scale_y_continuous(n.breaks =10)

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000, fill = "lightskyblue", color = "black") +
  scale_x_continuous(n.breaks=10) +
  scale_y_continuous(n.breaks=10) +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

Now it is time to interpret the plot.

• What do we see?
  • Minimum income seems to be around 9000, while maximum is around 54,000.
  • Most of the workers earn 15-25,000 TL
  • There are some people earning a lot and very few. Who are they? Interns? Managers?
• What is the shape?
  • We have long bins (bars) on the left side -> Right skewed (i know it does not make sense to me either, it is what it is)
  • Because the distribution is right-skewed, the mean will likely be larger than the median. The people on the right side are ruining the distribution -> outliers?
• What is the center?
  • The bulk of the workers seem to earn around the middle of 15,000 - 20,000 range.

How is income distributed?

Distribution of one continuous variable

Histogram

• Now that we understood our plot, we should make other people understand too by adding a title and a subtitle.

• Use the layer labs(title = "{your title}", subtitle = "{your subtitle}", x = "{x axis label}", y = "{y axis label}").

• When we write sentences we write them between quotation marks.

• Right: labs(title = "Monthly earnings...")

• Wrong: labs(title = Monthlyh earnings..)

• Right: labs(x = "Income", y = "N. of observations")

• Wrong: labs(x = Income, y = N. of observations)

How is income distributed?

Distribution of one continuous variable

Histogram

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000, fill = "lightskyblue", color = "black") +
  scale_x_continuous(n.breaks = 10) +
  scale_y_continuous(n.breaks = 10) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_classic()

How is income distributed?

Distribution of one continuous variable

Histogram

How is income distributed?

Distribution of one continuous variable

Density Plot

• Remember Statistics I?
  • We talked about random continuous random variables and their distributions.
  • We talked about density functions (uniform distribution, normal distribution etc.)

How is income distributed?

Distribution of one continuous variable

Density Plot

• \[ f(x) = \frac{1}{\sigma \sqrt{2\pi}} \exp\left(-\frac{(x-\mu)^2}{2\sigma^2}\right) \]

How is income distributed?

Distribution of one continuous variable

Density Plot

• Since income is a continuous random varaible, one alternative to use histogram is to use a density plot.
• In a density plot, instead of grouping the data into bins, R estimates a continuous curve that represents the distribution. (It basically calculates the density function and plots it).
• Lets see how income is distributed with a density plot.

How is income distributed?

Distribution of one continuous variable

Density Plot

We want to use our data named earnings and do ( %>% )

ggplot() visualization to it

Add x axis variable as income

Add density geometry: geom_density()

How is income distributed?

Distribution of one continuous variable

Density Plot

earnings %>%
  ggplot() +
  aes(x = income) +
  geom_density()

1

We want to use our data named earnings and do ( %>% )

2

Apply ggplot() visualization to it

3

Add x axis variable as income

4

Add density geometry: geom_density()

How is income distributed?

Distribution of one continuous variable

Density Plot

How is income distributed?

Distribution of one continuous variable

Density Plot

Immediately apply a theme you prefer by.

Well, before that, lets install a nice package to use a nice theme

We want to install the package ggthemes to use its themes

install.packages("ggthemes")

How is income distributed?

Distribution of one continuous variable

Density Plot

To be able use the themes from ggthemes package, we need to tell R to use the package

library(ggthemes)

We were here

earnings %>%
  ggplot() +
  aes(x = income) +
  geom_density()

How is income distributed?

Distribution of one continuous variable

Density Plot

Now we add a theme of your choice by starting to type theme_ on the right side of the suggestions you will see ggthemes, choose one from there.

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density() +
  theme_economist()

How is income distributed?

Distribution of one continuous variable

Density Plot

I don’t like the black color of the density line. Lets change it to “azure4” using the color argument inside geom_density(color = {"color you prefer"})

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density(color = "azure4") +
  theme_economist()

How is income distributed?

Distribution of one continuous variable

Density Plot

The width of the line seems a bit thin, can we increase it with linewidth argument inside geom_density(). By default it is set to geom_density(linewidth = 1).

geom_density(color = {"color you prefer"}, linewidth = {width number you prefer})

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density(color = "azure4", linewidth = 1.5) +
  theme_economist()

How is income distributed?

Distribution of one continuous variable

Density Plot

Label of the y axis, changed, because we are not showing counts anymore. We are showing density (or probability density). Our heights now represent probability density, not the number of workers; we should reflect this in y label.

The x and y values seem ok. we don’t need to use scale_x_continuous() and scale_y_continuous().

While our title is still fine, our subtitle could be better since we are not showing counts anymore. Lets change it to “The highest concentration of earnings lies between 15,000–25,000 TL; the distribution is right-skewed with a long upper tail”

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density(color = "azure4", linewidth = 1.5) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "The highest concentration of earnings lies between 15,000–25,000 TL; the distribution is right-skewed",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Probability Density"
  ) +
  theme_economist()

How is income distributed?

Distribution of one continuous variable

Density Plot

How is income distributed between genders?

Distribution of one continuous variable over a group

Density Plot

• Now, what if we wanted to see whether the income of the workers change based on their gender?

• Since we have gender for each observation, we can use this information as well.

• There are different ways to do this.

  • We can change the type of line based on gender.
  • We can change the color of the line based on gender.

How is income distributed between genders?

Distribution of one continuous variable over a group

Density Plot - change line color based on gender

This was our code and plot

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density(color = "azure4", linewidth = 1.5) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_economist()

How is income distributed between genders?

Distribution of one continuous variable over a group

Density Plot - change line color based on gender

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density(color = "azure4", linewidth = 1.5) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_economist()

• Now we want to change the color of the line, based on a variable.

• Can we change the color inside geom_density(color = "azure4") to geom_density(color = gender)?

• Try it yourself

How is income distributed between genders?

Distribution of one continuous variable over a group

Density Plot - change color type based on gender

earnings %>%  
  ggplot() + 
  aes(x = income) + 
  geom_density(color = gender, linewidth = 1.5) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_economist()

We get an error Error: object 'gender' not found.

This is because R thinks we give it a color name, like red, blue, or green. But gender is not a color name. It is a variable

Whenever we want to change something based on a variable, we do it inside aes(), not geom_.

This is why we set aes(x = income) instead of geom_density(x = income). Because x should be a varaible, not just one value.

So what we want to do is to go upstairs to aes(x = income), and add color = gender argument to it, seperating them with ,

How is income distributed?

Distribution of one continuous variable over a group

Density Plot

earnings %>%  
  ggplot() + 
  aes(x = income, color  = gender) + 
  geom_density(linewidth = 1.5) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_economist()

How is income distributed?

Distribution of one continuous variable over a group

Density Plot

• We added the gender information into our plot.

• Now we are no longer describing one distribution.

• We are comparing two distributions.

• The thing we show now is “Do earnings differ by gender?”.

• Our title and subtitle should reflect this.

• Change the title to “Monthly Earnings by Gender”.

  • Lets assume our audience know the data comes from Cilek Mobilya, Mamak (to shorten the title).

• Change the subtitle to “No substantial difference between male and female workers.”

How is income distributed?

Distribution of one continuous variable over a group

Density Plot

earnings %>%  
  ggplot() + 
  aes(x = income, color  = gender) + 
  geom_density(linewidth = 1.5) +
    labs(
    title = "Monthly Earnings by Gender",
    subtitle = "No substantial visual difference between male and female workers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_economist()

How is income distributed?

Distribution of one continuous variable over a group

Density Plot - change color type based on gender

Now, instead of the color, we could have changed the linetype based on gender.

Can you do it?

earnings %>%  
  ggplot() + 
  aes(x = income, linetype  = gender) + 
  geom_density(linewidth = 1.5) +
    labs(
    title = "Monthly Earnings by Gender",
    subtitle = "No substantial visual difference between male and female workers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_economist()

How is income distributed?

Distribution of one continuous variable over a group

Density Plot - change color type based on gender

How is income distributed?

Distribution of one continuous variable over a group

Histogram

Remember our histogram code?

earnings %>% 
  ggplot() +
  aes(x = income) +
  geom_histogram(binwidth = 1000, fill = "lightskyblue", color = "black") +
  scale_x_continuous(n.breaks = 10) +
  scale_y_continuous(n.breaks = 10) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_classic()

How is income distributed?

Distribution of one continuous variable over a group

Histogram

Instead of setting one color to fill the bins, lets fill the bins based on gender. While there, change the theme to a nicer one.

earnings %>% 
  ggplot() +
  aes(x = income, fill = gender) +
  geom_histogram(binwidth = 1000, color = "black",) +
  scale_x_continuous(n.breaks = 10) +
  scale_y_continuous(n.breaks = 10) +
    labs(
    title = "Monthly Earnings Distribution of Manufacturing Workers in Cilek Mobilya, Mamak",
    subtitle = "Most workers earn between 15,000–25,000 TL; distribution is right-skewed with a small number of high-income outliers",
    x = "Monthly Earnings (Turkish Lira, TL)",
    y = "Number of Workers"
  ) +
  theme_classic()

How is income distributed?

Distribution of one continuous variable over a group

Histogram

• So many things are wrong with this plot

• Use only one variable in histograms, if you need to make a group based distribution comparison use density plot.

• If your goal is to show “How many workers fall into each income range?”, histogram is the right plot (density plot is also ok btw, the interpretation differs slightly.).

  • y axis shows counts
  • You can say there are about 10 workers earning between .. - .. TL.

• If your goal is to show “Do the income distributions differ by gender?”, density plot is the right one

  • its shape makes us compare groups easier.
  • we have no overlap to worry about.
  • when we compare distributions, it is better to use a density plot.