Making a bar plot with stack and dodge, and keep the dodged bars touching one another

Question

I have data as follows:

library(ggplot2)
library(tidyr)
library(dplyr)

DT <- structure(list(value = structure(c(2, 3, 3, 4, 4, 5, 6, 6, 7, 
7, 8, 8, 9, 9, 1, 1, 2, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9
), label = NA_character_, class = c("labelled", "numeric")), 
    penalty = structure(c(0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 1, 
    0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1), label = "", class = c("labelled", 
    "numeric")), count = c(1L, 1L, 2L, 3L, 3L, 5L, 11L, 2L, 30L, 
    10L, 48L, 13L, 62L, 16L, 1L, 1L, 1L, 2L, 7L, 4L, 10L, 4L, 
    19L, 6L, 33L, 7L, 39L, 10L, 50L, 13L), type = c("Truth", 
    "Truth", "Truth", "Truth", "Truth", "Truth", "Truth", "Truth", 
    "Truth", "Truth", "Truth", "Truth", "Truth", "Truth", "Tax", 
    "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", 
    "Tax", "Tax", "Tax", "Tax", "Tax", "Tax"), x_label = c("2_Truth", 
    "3_Truth", "3_Truth", "4_Truth", "4_Truth", "5_Truth", "6_Truth", 
    "6_Truth", "7_Truth", "7_Truth", "8_Truth", "8_Truth", "9_Truth", 
    "9_Truth", "1_Tax", "1_Tax", "2_Tax", "3_Tax", "4_Tax", "4_Tax", 
    "5_Tax", "5_Tax", "6_Tax", "6_Tax", "7_Tax", "7_Tax", "8_Tax", 
    "8_Tax", "9_Tax", "9_Tax")), row.names = c(NA, -30L), class = c("data.table", 
"data.frame"))

   value penalty count  type x_label
 1:     2       0     1 Truth 2_Truth
 2:     3       0     1 Truth 3_Truth
 3:     3       1     2 Truth 3_Truth
...
28:     8       1    10   Tax   8_Tax
29:     9       0    50   Tax   9_Tax
30:     9       1    13   Tax   9_Tax

I tried to do a "bar plot with stack and dodge" as suggested in this link by Kent Johnson follows:

ggplot(DT, aes(x=x_label, y=count, fill=penalty)) +
  geom_bar(stat='identity') + labs(x='Value / Treatment') + 
  theme(legend.title = element_blank(), legend.position = c(0.1, 0.85))

However, I would like the chart to be more like the chart below (which is how I started out). Where at least the bars with the same value are together but have a different colour. If possible, also with the density function, the legend and the x-axis values.

Is there any way to do this?

EDIT: Suggestion of brianavery

some clarification of what variables you want to plot would be helpful, it seems like there is one less variable on the plot that you want to replicate? — brian avery, Jan 29 '21 at 15:18
@brianavery What I want to plot it already there, it is the first picture I added. The only thing I want to change is the way it looks. The way I want it to look is the second picture (of course while keeping the "stack"). — Tom, Jan 29 '21 at 15:20
if you take the answer from @tjebo but use `fill = as.factor(penalty)` for the fill to make it a factor and therefore discrete does that help with the color scale problem? I do think it will be very hard to get a density on there too with an approach using summarized data. — brian avery, Jan 29 '21 at 15:30
I put you suggestion under EDIT. It only seems to change the colours.. — Tom, Jan 29 '21 at 15:33

score 4 · Accepted Answer · answered Jan 29 '21 at 15:38

4

you can try

DT %>%
  mutate(value =as.character(value)) %>% 
  complete(crossing(value,type, penalty), fill = list(count = NA)) %>% 
  ggplot(aes(x= value, y=count, fill =  type)) +
  geom_col(data = . %>% filter(penalty==0), position = position_dodge(width = 0.9), alpha = 0.2) + 
  geom_col(data = . %>% filter(penalty==1), position = position_dodge(width = 0.9), alpha = 1)  +
  geom_tile(aes(y=NA_integer_, alpha = factor(penalty)))

answered Jan 29 '21 at 15:38

Roman

17,008
3
36
49

This great! Thank you so much! – Tom Jan 29 '21 at 15:40
Okay, I see why you chose factor 0 to be transparent.. I tried to change it, but then you cannot see the other colour any more. I have a feeling there is little to be done about that, or is there? The other way around would be exactly what I need. – Tom Jan 29 '21 at 15:42
1

interesting approach! Would you care adding this to the main related thread as well for better visibility? https://stackoverflow.com/questions/12715635/ggplot2-bar-plot-with-both-stack-and-dodge – tjebo Jan 29 '21 at 16:28
@tjebo thanks for the suggestion. I will do it in the the next days. – Roman Feb 02 '21 at 18:58
Please let me know - so I won't forget to upvote :) – tjebo Feb 02 '21 at 20:53

score 1 · Answer 2 · answered Jan 29 '21 at 14:46

The only way I see to get both stack and dodge but with the bars close together would be to use facets and to fake a continuous graph. Apologies still without real R console, using rdrr.io/snippets right now.

library(tidyverse)

DT <- structure(list(value = structure(c(2, 3, 3, 4, 4, 5, 6, 6, 7, 
7, 8, 8, 9, 9, 1, 1, 2, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9
), label = NA_character_, class = c("labelled", "numeric")), 
    penalty = structure(c(0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 1, 
    0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1), label = "", class = c("labelled", 
    "numeric")), count = c(1L, 1L, 2L, 3L, 3L, 5L, 11L, 2L, 30L, 
    10L, 48L, 13L, 62L, 16L, 1L, 1L, 1L, 2L, 7L, 4L, 10L, 4L, 
    19L, 6L, 33L, 7L, 39L, 10L, 50L, 13L), type = c("Truth", 
    "Truth", "Truth", "Truth", "Truth", "Truth", "Truth", "Truth", 
    "Truth", "Truth", "Truth", "Truth", "Truth", "Truth", "Tax", 
    "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", "Tax", 
    "Tax", "Tax", "Tax", "Tax", "Tax", "Tax"), x_label = c("2_Truth", 
    "3_Truth", "3_Truth", "4_Truth", "4_Truth", "5_Truth", "6_Truth", 
    "6_Truth", "7_Truth", "7_Truth", "8_Truth", "8_Truth", "9_Truth", 
    "9_Truth", "1_Tax", "1_Tax", "2_Tax", "3_Tax", "4_Tax", "4_Tax", 
    "5_Tax", "5_Tax", "6_Tax", "6_Tax", "7_Tax", "7_Tax", "8_Tax", 
    "8_Tax", "9_Tax", "9_Tax")), row.names = c(NA, -30L), class = c("data.table", 
"data.frame"))

DT %>%
ggplot(aes(x= as.character(value), y=count, fill = penalty, group = type)) +
  geom_col(position = "dodge") + 
  labs(x='Value / Treatment') + 
  facet_wrap(~ value, scale = "free_x", nrow = 1)+
  theme(legend.title = element_blank(), legend.position = c(0.1, 0.85),
panel.spacing = unit(0, "inch"))

This approach might make it tricky to produce a density curve, although the lates version of ggplot allows to do "cropped" density curves, this may also imitate continuity. Cannot try this out at the moment.

Thank you so much! Is there anything to be done about the colours? I tried adding `scale_fill_brewer(palette = "Set1") + ` but it says: `Error: Continuous value supplied to discrete scale` — Tom, Jan 29 '21 at 14:55
@Tom of course. factorise/categorise penalty - e.g. as.character(penalty) — tjebo, Jan 29 '21 at 15:35
It does change the colours now haha. But it does not give them a different colour (the bars that are next to each other always have the same colour). I would need four colours instead of two. — Tom, Jan 29 '21 at 15:38

score 0 · Answer 3 · answered Jan 29 '21 at 15:07

it seems easier to deal with the barchart part the way you have the data structured (as a summary), something like this:

ggplot(DT) +
  geom_col(aes(x=x_label, y=count, fill=as.factor(penalty)), position="dodge") + 
  labs(x='Value / Treatment') + 
  theme(legend.title = element_blank(), legend.position = c(0.1, 0.85))

making the variable you want to fill by a factor will make the color scale discrete and behave like I think you are after.

but I think the density will be difficult without the original data to plot (like a histogram).

can you use the original data to make a histogram and density? or provide an example of what it looks like?

Making a bar plot with stack and dodge, and keep the dodged bars touching one another

3 Answers3

Linked