5.5 Categorical variables (2+)

5.5.1 Data & Packages & functions

Data: Two or several categorical variables
Challenge: We need to summarize the data as to obtain frequencies (absolute or relative)
- Create long format dataframe that contains the frequencies of different category combinations
Packages & functions:
- tidyr and pivot_longer() function
- dplyr and functions such as summarize(), mutate() etc.

5.5.2 Graph

Here we’ll reproduce and maybe criticize as well as improve Figure 5.5 (Bauer and Clemm von Hohenberg 2020)
Questions:
- What does the graph show? What are the underlying variables (and data)?²⁹
- How many scales/mappings does it use? Could we reduce them?³⁰
- What do you like, what do you dislike about the figure? What is good, what is bad?
- What kind of information could we add to the graph (if any)?
- How would you approach a replication of the graph?

Figure 5.5: Several categorical variables

5.5.3 Lab: Data & Code

The code for Figure 5.5 is shown below (and creates Figure 5.6).
Learning objectives
- How to manipulate data first and visualize it thereafter
  - How to use pivot_longer (ggplot likes long format!)
- How to visualize unordered and ordered variables
- How to name scales and create manual ones
- How to create labels from data
  - Make sure the only thing that you might to add for the labels is gsub, i.e., now substantive changing of label names
- How to size text elements

We start by importing the original (unsummarised data). As you can see below we have categorical string variables.

# data_account_prevalence.csv
#data <- read_csv(sprintf("https://docs.google.com/uc?id=%s&export=download",
#                         "1xbffXai-HqWS2Q17KoB_MCCY7YOpskyH"))

data <- read_csv("data/data_account_prevalence.csv",
                 col_types = cols())

head(data)

account_email	account_fb	account_twitter	account_whatsapp
Yes	Yes	No	Yes
Yes	Yes	No	Yes
Yes	Yes, but I dont use it	No	Yes
Yes	Yes	No	Yes
Yes	No	No	Yes
Yes	Yes	Yes	Yes

Subsequently, we summarize/aggregate the data producing a dataframe that contains the percentage of people in each category (Yes, Yes, but inactive, No) across the four variables. It’s a good idea to call the data that builds the basis for our plot data_plot as to keep the original dataset data. Let’s go through the code below step by step:

# Creating plot data
  data_plot <- data %>% 
  
      pivot_longer(cols = account_email:account_whatsapp, # formerly gather
                   names_to = "variable", 
                   values_to = "value") %>%
  
      group_by(variable) %>%
  
      summarize(pct.Yes = mean(value == "Yes", na.rm=TRUE),
                pct.Inactive = mean(value == "Yes, but I dont use it", na.rm=TRUE),
                pct.No = mean(value == "No", na.rm=TRUE)) %>% 
  
      pivot_longer(cols = pct.Yes:pct.No, # formerly gather
                   names_to = "category", 
                   values_to = "value") %>% # only keep variables of interest
  
      mutate(category = factor(category, # Create factor for ordering
                               levels = c("pct.Yes", "pct.Inactive", "pct.No"), 
                               ordered = TRUE)) %>%
  
      mutate(value = round(value,2)) %>% 
      mutate(value = 100 * value)

# Change the labels and translate to english! (so we can direclty pull them out later)
  data_plot$category <- gsub("Inactive", "Yes, but inactive", data_plot$category)
  data_plot$category <- gsub("pct.", "", data_plot$category)
  
  data_plot$category <- factor(data_plot$category,
                               levels = c("Yes", "Yes, but inactive", "No"),
                               ordered = TRUE)
  data_plot$variable <- factor(data_plot$variable)
  
  data_plot$variable <- str_to_title(gsub("fb", "facebook", gsub("account_", "", 
                                          data_plot$variable)))
  data_plot

variable	category	value
Email	Yes	98
Email	Yes, but inactive	1
Email	No	1
Facebook	Yes	61
Facebook	Yes, but inactive	8
Facebook	No	30
Twitter	Yes	15
Twitter	Yes, but inactive	8
Twitter	No	77
Whatsapp	Yes	83
Whatsapp	Yes, but inactive	1
Whatsapp	No	16

Check out the variables in the data. Importantly, there is an ordered factor in there:

str(data_plot)

## tibble [12 x 3] (S3: tbl_df/tbl/data.frame)
##  $ variable: chr [1:12] "Email" "Email" "Email" "Facebook" ...
##  $ category: Ord.factor w/ 3 levels "Yes"<"Yes, but inactive"<..: 1 2 3 1 2 3 1 2 3 1 ...
##  $ value   : num [1:12] 98 1 1 61 8 30 15 8 77 83 ...

Now we have prepared the data we can plot it in Figure 5.6 (fairly easy!). Again let’s go through this step by step:

  # CHECK
# Creating the plot
ggplot(data_plot,aes(x = variable, y = value))+
  geom_bar(stat="identity", 
           width=0.7, 
           position = position_dodge(width=0.8),
           aes(fill = factor(variable),
               alpha=category)) +
  geom_text(position = position_dodge(width=0.8),
           aes(alpha=category,
               label = paste(value,"%", sep="")), 
           vjust=1.6, 
           color="black", 
           size=2) +
  
  scale_fill_discrete(name="Platform") +
  
  scale_alpha_discrete(name="Account",
                       range=c(1, 0.5)) + 
  xlab("Platforms")+
  ylab("Percentage (%)")+ 
  theme_light() +
theme(axis.text.x=element_text(angle=35,hjust=1,vjust=1, margin=margin(0.2,0,0.3,0,"cm")), 
      plot.title = element_text(hjust = 0.5),
      plot.margin = margin(0.5, 0.5, 0, 0.5, "cm"),      
      panel.grid.major.x = element_blank(),
      legend.title = element_text(size=10),
      legend.text = element_text(size=9))

Figure 5.6: Distribution of four categorical variables

Strictly speaking the coloring in Figure 5.6 would not be necessary as the platforms are already encoded on the x-Axis. Figure 5.6 uses 4 mappings (x, y, alpha/luminance, color) for three variables. Hence, we could also use a grayscale version of the graph that you see below in Figure 5.7.³¹

# Creating the plot
ggplot(data_plot,aes(x = variable, y = value))+
  geom_bar(stat="identity", 
           width=0.7, 
           position = position_dodge(width=0.8),
           aes(#fill = factor(variable),
               alpha=factor(category))) +
  geom_text(position = position_dodge(width=0.8),
           aes(alpha=factor(category),
               label = paste(value,"%", sep="")), 
           vjust=1.6, 
           color="black", 
           size=2) +
  

  
  scale_alpha_discrete(name="Account",
                       range=c(1, 0.5),
                       labels=c("Yes", "Yes, but inactive", "No")) + 
  
  scale_x_discrete(labels=str_to_title(gsub("fb", "facebook", 
                                            gsub("account_", "", unique(data_plot$variable))))) + 
  
  xlab("Platforms")+
  ylab("Percentage (%)")+ 
  theme_light() +
theme(axis.text.x=element_text(angle=35,hjust=1,vjust=1, margin=margin(0.2,0,0.3,0,"cm")), 
      plot.title = element_text(hjust = 0.5),
      plot.margin = margin(0.5, 0.5, 0, 0.5, "cm"),      
      panel.grid.major.x = element_blank(),
      legend.title = element_text(size=10),
      legend.text = element_text(size=9))

Figure 5.7: Distribution of four categorical variables

5.5.4 Exercise

Figure 5.8 different data but code that is very similar to Figure 5.6. Can you recreate it?

Load the summarized data (we’ll skip the data management steps).

# data_sharing_frequency_summarized.csv
#data_plot <- read_csv(sprintf("https://docs.google.com/uc?id=%s&export=download",
                              #"16oXege9RqvtIkBppZvkQNb4gl-lmSyoU"))
data_plot <- read_csv("www/data/data_sharing_frequency_summarized.csv")
head(data_plot)

variable	category	value
Email	Daily	7
Facebook	Daily	7
Twitter	Daily	7
Whatsapp	Daily	17
Email	Rarer	67
Facebook	Rarer	53

Try to recreate Figure 5.8 using the code from Figure 5.6. There is a mistake in Figure 5.8. Can you spot it?
How would we modify the code (data) if you want to show just 2 out of four variables or just 3 out of 5 categories?
How could we visualize a fourth categorical variable?
Here we used colors for differentiating. What would be an alternative way?

Figure 5.8: Distribution of four categorical variables

References

Bauer, Paul C, and Bernhard Clemm von Hohenberg. 2020. “Believing and Sharing Information by Fake Sources: An Experiment.”

Data: Data is four categorical variables with the same ordered categories. In the graph they are combined, i.e., the four variables are combined into two categorical variables: Platform with unordered categories and Account with 3 ordered categories.↩︎
We use both the x-scale and color for the same mapping namely platforms. This could be reduced.↩︎
One could also think of choosing colors that are also discernible when printed in grayscale instead of luminance.↩︎