Libraries

library(tidyverse)
library(cowplot)
library(gghighlight)
library(knitr)
library(MetBrewer)

Loading data

music <- read_csv("https://raw.githubusercontent.com/lescai-teaching/dataviz-rstudio/datasets/music_sales_history.csv",
                  col_names = c("format", "metric", "year", "records", "value"),
                  col_types = c(col_character(),
                                col_character(),
                                col_double(),
                                col_double(),
                                col_double()),
                  skip = 1)

Tidying the data

simplifying the column name

music$metric <- ifelse(
  music$metric == "Value (Adjusted)",
  "adjusted_value",
  music$metric
)

some data are duplicated or missing

music[duplicated(music),]

## # A tibble: 39 x 5
##    format             metric  year records value
##    <chr>              <chr>  <dbl>   <dbl> <dbl>
##  1 Paid Subscriptions Units   1973       1    NA
##  2 Paid Subscriptions Units   1974       1    NA
##  3 Paid Subscriptions Units   1975       1    NA
##  4 Paid Subscriptions Units   1976       1    NA
##  5 Paid Subscriptions Units   1977       1    NA
##  6 Paid Subscriptions Units   1978       1    NA
##  7 Paid Subscriptions Units   1979       1    NA
##  8 Paid Subscriptions Units   1980       1    NA
##  9 Paid Subscriptions Units   1981       1    NA
## 10 Paid Subscriptions Units   1982       1    NA
## # … with 29 more rows

Despite check, some data still seem to be duplicated, so when we group we’re just gonna choose the first value

music_tidy = music %>%
  filter(!is.na(value)) %>%
  pivot_wider(
    names_from = metric,
    values_from = value,
    values_fn = first
  )

the categories are pretty broad

unique(music_tidy$format)

##  [1] "CD"                                 "CD Single"                         
##  [3] "Cassette"                           "Cassette Single"                   
##  [5] "LP/EP"                              "Vinyl Single"                      
##  [7] "8 - Track"                          "Other Tapes"                       
##  [9] "Music Video (Physical)"             "DVD Audio"                         
## [11] "SACD"                               "Download Single"                   
## [13] "Download Album"                     "Kiosk"                             
## [15] "Download Music Video"               "Ringtones & Ringbacks"             
## [17] "Paid Subscriptions"                 "Limited Tier Paid Subscription"    
## [19] "On-Demand Streaming (Ad-Supported)" "Other Ad-Supported Streaming"      
## [21] "Other Digital"                      "Paid Subscription"                 
## [23] "SoundExchange Distributions"        "Synchronization"

In order to reduce them, we need to create a new category where we can simplify the existing media type descriptions:

music_tidy = music_tidy %>%
  mutate(
    media = case_when(
      grepl("CD", format) ~ "CD_DVD",
      grepl("DVD", format) ~ "CD_DVD",
      grepl("LP", format) ~ "Vinyl",
      grepl("Vinyl", format) ~ "Vinyl",
      grepl("Cassette", format) ~ "Cassette",
      grepl("Track", format) ~ "Vinyl",
      grepl("Download", format) ~ "Digital",
      grepl("Subscription", format) ~ "Digital",
      grepl("Ring", format) ~ "Digital",
      grepl("Streaming", format) ~ "Digital",
      grepl("Digital", format) ~ "Digital",
      grepl("Sync", format) ~ "Digital",
      grepl("SoundExchange", format) ~ "Digital",
      grepl("Tapes", format) ~ "Cassette",
      grepl("Physical", format) ~ "CD_DVD",
      grepl("Kiosk", format) ~ "Vinyl"
    )
  )

Stacked Area chart

The most common way to represent these data is a stacked area chart, to show both evolution and relative proportions of sales of different media types over the years.

music_tidy %>%
  group_by(media, year) %>%
  summarise(total_value = sum(adjusted_value, na.rm = TRUE)) %>%
  ggplot(aes(x=year, y=total_value, fill=media))+
  geom_area(position = "stack")

Beeswarm plot

We can also reproduce a beeswarm-like plot this way:

music_tidy %>%
  group_by(media, year) %>%
  summarise(total_value = sum(adjusted_value, na.rm = TRUE)) %>%
  ggplot(aes(x=year, y=0, colour=media, size = total_value))+
  geom_jitter(width = 0.005, alpha = 0.5)+
  ylim(-3,3)+
  scale_size(range = c(0.5, 10), name="Total Value by Year")+
  labs(
    x = "Year",
    y = "",
    colour = "Media type"
  )+
  theme(axis.text.y = element_blank())