tidyverse: using forcats to improve your ggplots

Note that although forcats is part of the “tidyverse”, it is not automatically loaded when you run library(tidyverse)

Handy forcats functions for ggplot2

Comparing followers of world religions

Source: https://en.wikipedia.org/wiki/List_of_religious_populations

I was looking for a simple dataset with count data for many items to demonstrate some basic forcats functions that are useful when creating plots.

religions = read_csv("https://raw.githubusercontent.com/acatlin/data/master/religions.csv", 
                  show_col_types = FALSE, col_names = FALSE) %>% 
    rename(religion = X1, followers = X2) %>% 
    mutate(millions_of_followers = followers/1000000.0) %>% 
  select(religion, millions_of_followers)

religions

# A tibble: 21 × 2
   religion                      millions_of_followers
   <chr>                                         <dbl>
 1 Christianity                                   2400
 2 Islam                                          1900
 3 Hinduism                                       1200
 4 Secular                                        1200
 5 Buddhism                                        506
 6 Chinese Traditional                             394
 7 Various Ethnic Religions                        300
 8 African traditional religions                   100
 9 Sikhism                                          26
10 Spiritism                                        15
# ℹ 11 more rows

1A: basic ggplot

Q: What are the most followed religions? A: Use ggplot to compare religious populations

You can also embed plots, for example:

religions %>% 
  ggplot(aes(x = religion, y = millions_of_followers)) + 
    geom_col(fill = "lightblue") + 
    labs(x = "religion", y = "millions of followers", 
           caption = "https://en.wikipedia.org/wiki/List_of_religious_populations") 

1B: How do I flip coordinates?

religions %>% 
  ggplot(aes(x = religion, y = millions_of_followers)) + 
    geom_col(fill = "lightblue") + 
    labs(x = "religion", y = "millions of followers", 
           caption = "https://en.wikipedia.org/wiki/List_of_religious_populations") + coord_flip()

2A: How do I change sort order?

Revised by: Andy Catlin

Q: How do we change the chart to show the most followed religions first? A: Use forcats::fct_reorder()

library(forcats) 

ggplot(religions, aes(x = fct_reorder(religion, millions_of_followers), 
                      y = millions_of_followers)) + 
  geom_col(fill = "lightblue") + 
  labs(x = "religion", y = "millions of followers", 
      caption = "https://en.wikipedia.org/wiki/List_of_religious_populations") + 
  coord_flip()

2B: How do I combine less frequently used categories?

Q: How do we combine the less-followed religions into a single group? A: Use forcats::fct_other()

top5 = unlist(select(head(arrange(religions, desc(millions_of_followers)), 5), religion))

religions %>% 
  mutate(religion = fct_other(religion, keep = top5, other_level = "Other religions")) %>%     
  ggplot(aes(x = fct_reorder(religion, millions_of_followers), y = millions_of_followers)) +
    geom_col(fill = "lightblue") + 
    labs(x = "religion", y = "millions of followers", 
        caption = "https://en.wikipedia.org/wiki/List_of_religious_populations") + 
        coord_flip()

2C: Adding a title

Reference: https://www.geeksforgeeks.org/ggplot2-title-and-subtitle-with-different-size-and-color-in-r/

religions %>% 
  mutate(religion = fct_other(religion, keep = top5, other_level = "Other religions")) %>%     
  ggplot(aes(x = fct_reorder(religion, millions_of_followers), y = millions_of_followers)) +
    geom_col(fill = "lightblue") + 
    labs(x = "religion", y = "millions of followers", 
         title = "Most Popular Religions",
         subtitle = "[2021]",
         caption = "https://en.wikipedia.org/wiki/List_of_religious_populations") +
    theme(plot.title = element_text(size = 18, color = "blue"),
        plot.subtitle = element_text(size = 14, color = "gold")) +
          coord_flip()

Tabular Data

religions |> 
  gt(rowname_col = "religion") |>
  tab_header(
              title = "Most popular religions",
              subtitle = md("**2021**")) |>
  tab_source_note(
           source_note = md("https://en.wikipedia.org/wiki/List_of_religious_populations")) |>
  opt_table_font(font = google_font("Montserrat"), weight = 500)

Most popular religions
2021
	millions_of_followers
Christianity	2400
Islam	1900
Hinduism	1200
Secular	1200
Buddhism	506
Chinese Traditional	394
Various Ethnic Religions	300
African traditional religions	100
Sikhism	26
Spiritism	15
Judaism	15
Baháʼí	7
Jainism	4
Shinto	4
Cao Dai	4
Zoroastrianism	3
Tenrikyo	2
Animism	2
Neo-Paganism	1
Unitarian Universalism	1
Rastafari	1
https://en.wikipedia.org/wiki/List_of_religious_populations

Findings and Recommendations

To use the terminology of descriptive analytics (vs. predictive analytics), there is a single measure (millions of followers) across a single level of a single dimension (religion). Suppose we were able to find counts of religion data every 10 years for the past 200 years, by continent.

Two useful patterns of analysis in descriptive analytics are relative contribution and changes over time.

Relative contribution: What is the percent of the total that each religion represents (overall? by continent?) Changes over time: How did the counts (and percentages) of different relgions change over time (overall? by continent?)

What other measures might be interesting (e.g. by age group)

How would you represent the information in a table or a chart?

Would you be able to forecast religion counts (by continent) into the future?

etc.