ggplot2 - Stacked bar with different width

Question

I want to produce a reverse pyramid graph where the bar stacked on each other with different width.

1st, I have a stacked bar chart as below code sample

library(dplyr)
library(ggplot2)
sample <- data_frame(x=c(1, 1, 1, 1, 2, 2, 2, 2),
                     y=c(5,10,15, 20, 10, 5, 20, 10),
                     w=c(1, 2, 3, 4, 1, 2, 3, 4),
                     group=c("a", "b", "c", "d", "a", "b", "c", "d"))

ggplot() +
    geom_bar(data=sample,
             aes(x=x,y=y,group=group, fill=group),
             stat="identity", position=position_stack())

Then I added the width to aes so the one with lower w value will be smaller while they still stacked on each other. However, the bars didn't stack with warnings.

ggplot() +
geom_bar(data=sample,
         aes(x=x,y=y,group=group, fill=group, width=w/5),
         stat="identity", position=position_stack())

Warning: Ignoring unknown aesthetics: width
Warning message:
position_stack requires non-overlapping x intervals 

Any helps on make bar plot stacked or ideas on different plot type that can cover similar concepts would be highly appreciated. Thanks!

Answer 1

It's a bit of a hack.

I'll use geom_rect() instead of real columns. As such I need to create data.frame() with precalculated positions for rectangle boundaries.

df_plot <- sample %>% 
  arrange(desc(group)) %>% # Order so lowest 'groups' firtst
  group_by(x) %>% 
  mutate(yc = cumsum(y), # Calculate position of "top" for every rectangle
         yc2 = lag(yc, default = 0) ,# And position of "bottom"
         w2 = w/5) # Small scale for width


# Plot itself

ggplot(df_plot) +
  geom_rect(
    aes(xmin = x - w2 / 2, xmax = x + w2 / 2,
        ymin = yc, ymax = yc2,
        group = group, fill=group))

Resulting plot:

Answer 2

A rather lengthy version with ribbons

library(dplyr)
library(ggplot2)
sample <- data_frame(x=c(1, 1, 1, 1, 2, 2, 2, 2),
                     y=c(5,10,15, 20, 10, 5, 20, 10),
                     w=c(1, 2, 3, 4, 1, 2, 3, 4),
                     group=c("a", "b", "c", "d", "a", "b", "c", "d"))

# make factors for non-numeic items
sample$x <- factor(sample$x)
sample$group <- factor(sample$group)

# calcualte cumulative sums
sample2 <- sample %>%
  group_by(x) %>%
  arrange(desc(group)) %>%
  mutate(ycum=cumsum(y)) %>%
  ungroup()  %>%
  select(x, group, ycum, w) 

# Ffor each point, make another row lagged  
sample2lead <- sample2 %>%
  group_by(x) %>%
  mutate(ycum = lag(ycum, default=0), w=lag(w, default=max(sample2$w))) %>%
  ungroup() %>%
  select(x, group, ycum, w) 

# combine   
combined <- bind_rows(sample2, sample2lead) %>%
  arrange(x, ycum, desc(group))


# plot a ribbon forming trapezoids
ggplot() +
  geom_ribbon(data=combined,
             aes(x=ycum, ymin=-w/2, ymax=w/2, fill=group)) +
  coord_flip() +
  facet_grid(~x)