Bar charts connected by lines / How to connect two graphs arranged with grid.arrange in R / ggplot2

Question

At Facebook research, I found these beautiful bar charts which are connected by lines to indicate rank changes:

https://research.fb.com/do-jobs-run-in-families/

Answer 1

Here's a pure ggplot2 solution, which combines the underlying data frames into one & plots everything in a single plot:

Data manipulation:

library(dplyr)    
bar.width <- 0.9

# combine the two data sources
df <- rbind(state1 %>% mutate(source = "state1"),
            state2 %>% mutate(source = "state2")) %>%

  # calculate each state's rank within each data source
  group_by(source, state) %>%
  mutate(state.sum = sum(value)) %>%
  ungroup() %>%
  group_by(source) %>%
  mutate(source.rank = as.integer(factor(state.sum))) %>%
  ungroup() %>%

  # calculate the dimensions for each bar
  group_by(source, state) %>%
  arrange(type) %>% 
  mutate(xmin = lag(cumsum(value), default = 0),
         xmax = cumsum(value),
         ymin = source.rank - bar.width / 2,
         ymax = source.rank + bar.width / 2) %>% 
  ungroup() %>%

  # shift each data source's coordinates away from point of origin,
  # in order to create space for plotting lines
  mutate(x = ifelse(source == "state1", -max(xmax) / 2, max(xmax) / 2)) %>%
  mutate(xmin = ifelse(source == "state1", x - xmin, x + xmin),
         xmax = ifelse(source == "state1", x - xmax, x + xmax)) %>%

  # calculate label position for each data source
  group_by(source) %>%
  mutate(label.x = max(abs(xmax))) %>%
  ungroup() %>%
  mutate(label.x = ifelse(source == "state1", -label.x, label.x),
         hjust = ifelse(source == "state1", 1.1, -0.1))

Plot:

ggplot(df, 
       aes(x = x, y = source.rank,
           xmin = xmin, xmax = xmax, 
           ymin = ymin, ymax = ymax,
           fill = type)) +
  geom_rect() +
  geom_line(aes(group = state)) +
  geom_text(aes(x = label.x, label = state, hjust = hjust),
            check_overlap = TRUE) +

  # allow some space for the labels; this may be changed
  # depending on plot dimensions
  scale_x_continuous(expand = c(0.2, 0)) +
  scale_fill_manual(values = fill) +

  theme_void() +
  theme(legend.position = "top")

Data source (same as @camille's):

set.seed(1017)

state1 <- data_frame(
  state = rep(state.name[1:5], each = 3),
  value = floor(runif(15, 1, 100)),
  type = rep(c("state", "local", "fed"), times = 5)
)

state2 <- data_frame(
  state = rep(state.name[1:5], each = 3),
  value = floor(runif(15, 1, 100)),
  type = rep(c("state", "local", "fed"), times = 5)
)