Tired of Pie Charts !? Let’s make a Tree Map.

Tree maps are useful alternatives for the presentation of proportions.

Data Reading

For this first blog post, we are going to use Tidytuesday(Mock 2021) Week 34 Dataset.

The data comes from SpeechInteraction.org and was shared by Sara Stoudt.

It is the records of Star Trek characters interactions with Voice User Interfaces(VUIs) such as Apple Siri, Google Assistant, Amazon Alexa.

A complete description of the dataset is available at : https://github.com/rfordatascience/tidytuesday/tree/master/data/2021/2021-08-17.

First of all, let’s import tidyverse package :

library(tidyverse) # Import ggplot2 and friends for data wrangling 

Secondly, let’s read the data :

computer <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-08-17/computer.csv')

Here is what the data looks like.

head(computer, 2)

name	char	line	direction	type	pri_type	domain	sub_domain	nv_resp	interaction	char_type	is_fed	error	value_id
102	Tasha	Battle bridge.	The doors snap closed and the lift moves. Riker looks Tasha over waiting then:	Statement	Statement	IoT	Turbolift	TRUE	Battle bridge.	Person	TRUE	FALSE	255
102	Beverly	Show me the results of Captain Picard’s most recent physical examination.	The screen promptly BEGINS TO FLASH UP PRINTED INFORMATION followed by X-Ray type shots etc. Beverly studies it for awhile.	Command	Command	InfoSeek	NA	TRUE	Show me the results of Captain Picard’s most recent physical examination.	Person	TRUE	FALSE	345

Data Wrangling

We won’t spend a lot of time wrangling data to try to bring out some specific pattern. We are going to simply count the characters which interact the more with VUIs.

characters <- computer %>% 
  count(char, sort= T) %>%   
  # Categorize  the  characters that will be useful for our visualization
  mutate(char = factor(char)) 

Let’s look at, the 6 characters which interact the most with VUIs.

head(characters)

char	n
Computer Voice	598
Geordi	320
Picard	266
Data	235
Riker	150
Beverly	121

The results are logical because Geordi is an engineer in The Star Trek Saga.

Graphic

For this section, we need to import some extra libraries:

• treemapify(Wilkins 2021)
• paletteer(file. 2021)
• ggtext
• ragg

library(treemapify) # For geom_treemap and friends
library(paletteer) # For color palette 
library(ggtext) # For customize text (used in this script with element_markdown)
library(ragg) # For the device for save the plot

So why a Tree Map, instead of a Pie chart for proportions Visualization?

Pie charts are generally really cool to look at. But in some cases, like ours, there is too many individual data, and they are so close in magnitude that the message that we try to spread with a pie chart is biased.

In those cases, I use Tree Maps, which are pretty useful for the presentation of proportions.

A tree map shows the attribute of a cardinally scaled variable as nested rectangles. The size and order of the rectangles are calculated so that, with preset outer dimensions, the large rectangle is completely filled and the areas of the individual rectangles correspond to the size of the variables.(Rahlf 2017)

characters %>%
  ggplot(aes(fill = char,area = n)) + 
  geom_treemap(color = "black", size = 1) + 
  # We won't use legends because we 'ill annotate each square 
  # with the character name and the number of times it interacted 
  theme(legend.position = "none") 

At this step, we just mapped, each area with the number of interactions of a character and filled the area with the character.

Next step, let’s annotate each area with the name of the matching character using the function geom_treemap_text of treemapify package.

This function takes several arguments especially:

• family for the font family
• fontface for the font face
• colour for the font color
  
• place for the place inside the box where the text is placed.
• grow which is is very important, because if TRUE, the text will be grown as well as shrunk to fill the box.

characters %>%
  # Add in the mapping, label
  ggplot(aes(fill = char,area = n,label = glue::glue(" {char} \n ({n})"))) + 
  geom_treemap(color = "black", size = 1) + 
  geom_treemap_text(family = "Lato Black",fontface = "italic",
                    colour = "white", place = "centre",
                    grow = TRUE) + 
  theme(legend.position = "none") 

It began to look pretty good, but let’s customize it a little bit.

For that, I will fill the areas with a different color palette. The challenge would be to find a color palette with more 25 colors.

Fortunately for us, colorRampPalette() R function provide a tools to interpolate a set of given colors to create new color palettes.

Cédric Scherer published a trick about that.

Let’s use it. For color palette, I choose Prism from rcartoclor package. We can also access this palette within paletteer package.

paletteer_d("rcartocolor::Prism",12) %>% 
  # We visualize the colors with show_col from scales pkg 
  scales::show_col()

Now, we extend it.

extended_palette <- colorRampPalette(paletteer_d("rcartocolor::Prism",12)) 
# Let's visualize 49 interpolated colors et
extended_palette(49) %>% 
  scales::show_col(cex_label = .55, ncol = 7)

Finally we can use it, and set a little bit more the theme.

plot <- characters %>%
  ggplot(aes(fill = char,area = n, label = glue::glue(" {char} \n ({n})"))) + 
  geom_treemap(color = "black", size = 1) + 
  geom_treemap_text(family = "Lato Black",fontface = "italic", colour = "white", place = "centre",
                    grow = TRUE) + 
  labs(title = "Which characters interact the most with VUIs\n in the Star Trek Saga?",
       caption = "Data from ***SpeechInteraction.com*** and shared by Sara Stoudt.<br/>
       Tidytuesday Week-34 2021.") + 
  scale_fill_manual(values = extended_palette(nrow(characters))) + 
  theme(text =element_text(family = "Lato"),
        plot.background = element_rect(fill = "grey95"),
        panel.spacing = unit(2.5, units = "cm"),
        plot.title = element_text(family = "Lato Black",size = rel(2.5), hjust = .5, margin = margin(t = 15,b = 10)),
        plot.caption = element_markdown(color = "black", size = rel(1.2), margin = margin(t = 20,b = 10)),
        legend.position = "none"
  )

Et voilà!