R Intensive Project

On this page

• Data Exploration on Soccer Players
• Load the relevant libraries
• Let’s look at the dataset first
  • Issues with dataset
  • Fix the dataset
• Distribution of height (and age) among different positions
  • Basic Box plot
    • Violin Plot
  • Adding age as another variable
    • Overlapping layers
    • (Somewhat) Informative layer plot
  • Heat map
    • Simple heat maps
    • More informative heatmap
• Distribution of names with special characters across countries
  • World map

Tip

Plots will be images or interactive plots.
Note: images can be clicked on to zoom in on the plot itself
Have fun! :)

Data Exploration on Soccer Players

Load the relevant libraries

library(dplyr)
library(ggplot2)
library(plotly)
library(tidyr)
library(knitr)

Let’s look at the dataset first

players_2024 <- read.csv("../../../../data/Players2024.csv") 

……..Loading data……..

What type of data exists in our dataset?

Headers
name
birth_date
height_cm
positions
nationality
age
club

\(\boldsymbol{Peak\text{ } at \text{ } data}\)

name	birth_date	height_cm	positions	nationality	age	club
James Milner	1986-01-04	175	Midfield	England	38	Brighton and Hove Albion Football Club
Anastasios Tsokanis	1991-05-02	176	Midfield	Greece	33	Volou Neos Podosferikos Syllogos
Jonas Hofmann	1992-07-14	176	Midfield	Germany	32	Bayer 04 Leverkusen Fußball
Pepe Reina	1982-08-31	188	Goalkeeper	Spain	42	Calcio Como
Lionel Carole	1991-04-12	180	Defender	France	33	Kayserispor Kulübü
Ludovic Butelle	1983-04-03	188	Goalkeeper	France	41	Stade de Reims

\(\boldsymbol{Summary\text{ } of \text{ } data}\)

	name	birth_date	height_cm	positions	nationality	age	club
	Length:5935	Length:5935	Min. : 17	Length:5935	Length:5935	Min. :15.0	Length:5935
	Class :character	Class :character	1st Qu.:178	Class :character	Class :character	1st Qu.:22.0	Class :character
	Mode :character	Mode :character	Median :183	Mode :character	Mode :character	Median :25.0	Mode :character
			Mean :183			Mean :25.5
			3rd Qu.:188			3rd Qu.:29.0
			Max. :206			Max. :42.0

Issues with dataset

Caution

Visualising height_cm against positions shows issues with dataset

# Standard box plots (error with missing)
ggplot(players_2024,
       aes(x = positions,
           y = height_cm)) +
  geom_boxplot()

Fix the dataset

Here we will filter out “bad” values

players <-  players_2024 %>% 
  filter(positions != "Missing", height_cm > 100)

Note

NOTE: this is a specific filter command for our case,
it is common to check for na, blank etc etc.
Always check your dataset or adopt good practices when making dataset(csv).

Distribution of height (and age) among different positions

Basic Box plot

We first look at distribution of the height of players across different positions.

# Boxplot with fixed dataset
p1 <- ggplot(players,
         aes(x = positions,
             y = height_cm)) +
    geom_boxplot()
ggplotly(p1)

Tip

First Interactive Plot!!!
Interact with the plot to get IQR information about plots

Violin Plot

We fancify the basic box plot by surrounding the box plot with a violin plot

# plot simple violin plot
ggplot(players, 
         aes(x = positions,
             y = height_cm)) +
    geom_violin(aes(fill=positions))+
    geom_boxplot(width=0.1)

Adding age as another variable

What if we just added the age in as another box plot (layered box plot). While this will tell us the distribution as well, it is rather unappealing and the box plots are squeezed as a result of the unnecessarily long scales.

# Include all 
ggplot(players,
       aes(x = positions,
           y = height_cm)) +
  geom_boxplot() +
  geom_boxplot(
    mapping = aes(y = age), data=players
  ) +
  scale_y_continuous(
    "height in centimetres", 
    sec.axis = sec_axis(~ . * 1.20, name = "age")
  )

Overlapping layers

What if we added age points surrounding the box-plot to get the distribution of ages. It is rather messy because there is far too much (age) data for the same heights

#Can i inform the audience about age? 
ggplot(players, 
       aes(x = positions,
           y = height_cm)) +
  geom_jitter(aes(colour=age))+
  geom_boxplot(width=0.1)

NoteSidenote

NOTE: I did attempt to add a distribution around the violin plot
with color gradient but it seems to more difficult that i thought
because the violin plot does not show data points but a distribution.
There are some stack overflow discussion on gradient colors in violin plots
but they are more complicated than a few hours debugging.

(Somewhat) Informative layer plot

Instead of plotting age as it is, let’s get the average age per height.
Lets check if it is sensible to do so.

#Collect for each position, the heights and find the average age per height 
#for each positions
height_age <- players %>% 
  group_by(height_cm,positions) %>%  
  arrange(desc(age)) %>% 
  summarise(mean_age=mean(age),
            n = n())

……..Collecting data……..

\(\boldsymbol{Peak\text{ } at \text{ } data}\)

height_cm	positions	mean_age	n
160	Midfield	25.00000	1
161	Attack	20.00000	1
163	Attack	30.00000	1
163	Defender	23.50000	2
164	Attack	21.33333	3
164	Midfield	28.50000	2

\(\boldsymbol{Summary \text{ } of \text{ } filtered \text{ } data}\)

	height_cm	positions	mean_age	n
	Min. :160.0	Length:141	Min. :20.00	Min. : 1.00
	1st Qu.:175.0	Class :character	1st Qu.:24.66	1st Qu.: 6.00
	Median :185.0	Mode :character	Median :25.36	Median : 34.00
	Mean :184.1		Mean :25.61	Mean : 42.07
	3rd Qu.:193.0		3rd Qu.:26.22	3rd Qu.: 65.00
	Max. :206.0		Max. :34.00	Max. :166.00

\(\boldsymbol{Interactive \text{ } Plot \text{ } of \text{ } mean\_age \text{ } against \text{ } height \text{ } with \text{ } count \text{ } of \text{ } players \text{ } given}\)

• Red: more than 100 players
• Orange: more than 10 but less than 99
• Grey: less than 10
• Number labels: Number of players of the same height

#feel of data
p2 <- ggplot(height_age,
          aes(x = height_cm,
              y = mean_age)) +
  geom_point()+
  geom_point(data = subset(height_age, n >= 100), color = "red")+
  geom_point(data = subset(height_age,  n>= 10 & n <= 99), color = "orange")+
  geom_point(data = subset(height_age, n>=0 & n <=9), color = "grey")+
  geom_text(aes(label = n), vjust = 1.2, size = 3)
ggplotly(p2)

We bring the violin plot (with the box plot within) back and
now add the mean age as coloured points

# Violin plot with box plot 
ggplot(players, 
       aes(x = positions,
           y = height_cm)) +
  geom_violin()+
  geom_boxplot(width=0.2)+
  geom_point(
    data= height_age,
    aes(x = positions, y = height_cm, colour =mean_age),
    size = 4
  ) + 
  scale_colour_viridis_c(option = "plasma")

Area for improvement: Ideally, we can achieve a violin plot with colour gradient
of the mean age instead of coloured points.🙃

Heat map

Let’s drop the whole box or violin plots. What about a heat map?

Simple heat maps

We could just plot a simple heatmap of height against age where positions
are shown in colours. However, positions are overlapping themselves.

# Instead of a violin plot and keeping positions as "1 axis", 
# what about a heatmap
#Ugly heatmap
ggplot(players,
       aes(x = age,
           y = height_cm)) + 
  geom_tile(aes(fill=positions),alpha=0.5)

Instead, we separate out the heatmaps into the individual positions. While this is more informative, the colours struggle to exactly show the distribution.

#Separated but ugly heatmap (no need show)
ggplot(players, 
          aes(x = age, 
              y = height_cm)) + 
  geom_tile(aes(fill=positions),alpha=0.7) + 
  facet_wrap(vars(positions))

More informative heatmap

Since positions is already relayed by the separation of the heatmaps, we can
instead keep the same colour scheme for the distribution,
We then get this more informative map of distribution of height against age

# Decent heat map that shows concentration of points
ggplot(players, aes(x = age, y = height_cm)) +
  geom_bin2d(aes(fill = after_stat(count)), binwidth = c(1, 2)) +  # 1-year x 2-cm bins
  facet_wrap(vars(positions))+
  scale_fill_viridis_c(trans = "sqrt", name = "Count")

Visually appealing heatmap

Hate the “0” data around the original data?
Fill up the 0s around the original data and get yourselves a more appealing
heatmap 🤗

Distribution of names with special characters across countries

We played around with positions, age, height.
What about other data in the dataset?
Could we do something with name and nationality?

World map

Some names of our players have special characters outside of the 26 characters (A-Z)
What if we looked at the distribution of players that have special characters
in their names?

players_non_english <- players %>%
    filter(grepl("[^A-Za-z '\\-]", name))

……..Extracting data……..

\(\boldsymbol{Peak\text{ } at \text{ } data}\)

name	birth_date	height_cm	positions	nationality	age	club
Raúl Albiol	1985-09-04	190	Defender	Spain	39	Villarreal Club de Fútbol S.A.D.
Jesús Navas	1985-11-21	170	Defender	Spain	38	Sevilla Fútbol Club S.A.D.
Rúben Fernandes	1986-05-06	187	Defender	Portugal	38	Gil Vicente Futebol Clube
Iván Cuéllar	1984-05-27	187	Goalkeeper	Spain	40	Real Club Deportivo Mallorca S.A.D.
João Moutinho	1986-09-08	170	Midfield	Portugal	38	Sporting Clube de Braga

……..Collecting data as needed……..

#Taken from exemplar 
players_cty <- players_non_english %>% 
  mutate(nationality = case_match(nationality,
                                  "England" ~ "United Kingdom",
                                  "Türkiye" ~ "Turkey",
                                  "Cote d'Ivoire" ~ "Ivory Coast",
                                  "Northern Ireland" ~ "United Kingdom",
                                  "Wales" ~ "United Kingdom",
                                  .default = nationality))

# Make the country count
countries <- players_cty %>%
  group_by(nationality) %>%
  summarise(n = n()) %>% 
  arrange(desc(n))

ImportantCREDIT

This part of the work (plotting world map) was adapted from exemplar for our data

\(\boldsymbol{Peak\text{ } at \text{ } data}\)

nationality	n
Spain	219
Turkey	212
Portugal	90
France	89
Argentina	56
Germany	46
Denmark	29
Ivory Coast	20
Colombia	19
Brazil	18
Serbia	17
Belgium	16
Uruguay	16
Iceland	14
Mali	13
Norway	13
Hungary	12
Netherlands	12
Italy	11
Venezuela	11

\(\boldsymbol{World\text{ } map \text{ } distribution}\)

# World map plots
countries %>% 
  plot_ly(type = "choropleth", 
          locations = countries$nationality, 
          locationmode = "country names", 
          z = countries$n) %>%
  colorbar(title = "# of Players")