facets

faceted plots (or small multiple plots) are a way to divide your data up by a categorical variable. Facets are "not a geom, but rather a way of organizing a series of geoms" (Kieran Healy).

facets: think about the comparison!

GDP/pc development by continent.

In ggplot, we use the facet_wrap() building block to specify the faceting variable(s).

p <- ggplot(data = gapminder, 
            mapping = aes(x = year, 
                          y = gdpPercap)) + 
  geom_line(color="gray70", 
            aes(group = country)) + # recall we need to map group to country
  facet_wrap(~ continent, # "~" 
             ncol = 5) # how many columns?

facets: think about the comparison!

print(p)

facets: all elements

library(tidyverse)
library(gapminder)
p <- ggplot(data = gapminder, 
            mapping = aes(x = year, 
                          y = gdpPercap)) + 
  geom_line(color="gray70", aes(group = country)) +
  # add smoother
  geom_smooth(size = 1.1, method = "loess", se = FALSE) +
  # log y axis (could've also wrapped y=log(gdpPercap) in aes() above)
  scale_y_log10(labels=scales::dollar) +
  # facet command
  facet_wrap(~ continent, ncol = 5) +
  # labels and appearance tweaks
  labs(x = "Year",
       y = "GDP per capita",
       title = "GDP per capita on Five Continents") +
  theme_bw() +
  theme(axis.text.x = element_text(size = 5))

facets: all elements

print(p)

facets: more applications

Recall our example: relationship between GDP per capita and population in Asia.

gdp_pop_plot <- ggplot(gapminder %>% filter(continent == "Asia"), 
                       aes(x = log(pop), 
                           y = log(gdpPercap))) +
  geom_point(alpha = 0.5, size = 2) +
  geom_smooth(method = "lm")
print(gdp_pop_plot)

facets: more applications

Plot regression line by country (without facets)

gdp_pop_plot <- ggplot(gapminder %>% filter(continent == "Asia"), 
                       aes(x = log(pop), 
                           y = log(gdpPercap))) +
  geom_point(aes(color = country), 
             alpha = 0.5, size = 2) +
  geom_smooth(aes(fill = country, color = country),
              method = "lm") +
  theme(legend.position = "none")

facets: more applications

facets: more applications

Prior plot useful for MoMA, but not for the data analyst. How to do better?

facets!

gapminder_asia <- gapminder %>% 
  filter(continent == "Asia")
gdp_pop_plot <- ggplot(gapminder_asia, 
                       aes(x = log(pop), 
                           y = log(gdpPercap))) +
  geom_point(alpha = 0.5, size = 1) +
  geom_smooth(method = "lm", size = 0.7) + 
  facet_wrap(~ country, scales = "free") + # scales = "free" to vary axis limits +
  theme_bw() +
  theme(axis.text = element_text(size = 4), 
        strip.text = element_text(size = 6))

facets: more applications

facets: order by summary statistic

library(forcats) # useful to reorder factors or ordered categorical variables
gapminder_asia <- gapminder %>% 
  filter(continent == "Asia") %>% 
  # do all data manipulation by country
  group_by(country) %>% 
  # extract beta coefficient from reg of GDP on pop 
  mutate(beta = coef(lm(log(gdpPercap) ~ log(pop)))[2]) %>%  
  # remove country grouping 
  ungroup() %>% 
  # sort "country" variable by beta
  mutate(country_order = fct_reorder(country, beta))
head(gapminder_asia, 5)

## # A tibble: 5 x 8
##       country continent  year lifeExp      pop gdpPercap        beta
##        <fctr>    <fctr> <int>   <dbl>    <int>     <dbl>       <dbl>
## 1 Afghanistan      Asia  1952  28.801  8425333  779.4453 -0.03799305
## 2 Afghanistan      Asia  1957  30.332  9240934  820.8530 -0.03799305
## 3 Afghanistan      Asia  1962  31.997 10267083  853.1007 -0.03799305
## 4 Afghanistan      Asia  1967  34.020 11537966  836.1971 -0.03799305
## 5 Afghanistan      Asia  1972  36.088 13079460  739.9811 -0.03799305
## # ... with 1 more variables: country_order <fctr>

facets: order by summary statistic

facets: Exercise

Plot the relationship between gdpPercap and lifeExp in the Americas, faceted by country.

Bonus: sort country by the direction + strength of the relationship between gdpPercap and lifeExp

What is surprising?

facets: Exercise Solution

gapminder_americas <- gapminder %>% 
  filter(continent == "Americas") %>% 
  group_by(country) %>% 
  mutate(beta = coef(lm(lifeExp ~ log(gdpPercap)))[2]) %>% 
  ungroup() %>% 
  mutate(country_order = fct_reorder(country, beta))
gdp_lifeexp_americas <- ggplot(gapminder_americas, 
                       aes(x = log(gdpPercap), 
                           y = lifeExp)) +
  geom_point(alpha = 0.5, size = 1) +
  geom_smooth(method = "lm", size = 0.7) + 
  facet_wrap(~ country_order,  # facet by "country_order"!
             scales = "free") + # scales = "free" to vary axis limits + 
  theme_bw() +
  theme(axis.text = element_text(size = 4), 
        strip.text = element_text(size = 6))

facets: Exercise Solution