How do I make a barplot in R with a table of boolean categories as the x axis label?-CodePudding

I have a dataframe of counts of cases categorised by 5 attributes, each attribute is either TRUE, FALSE or NA. I'm trying to find a way of presenting the counts of the different combinations of attributes as a plot in R. Here is some data (the real data has a few more attributes):

> df <- data.frame( attibute_A = c(T,F,T,NA,F), attribute_B = c(F,T,T,NA,T), attribute_C = c(T,T,F,F,F), attribute_D = c(T,T,NA,NA,F), count=c(100,55,34,12,3))
> df
  attibute_A attribute_B attribute_C attribute_D count
1       TRUE       FALSE        TRUE        TRUE   100
2      FALSE        TRUE        TRUE        TRUE    55
3       TRUE        TRUE       FALSE          NA    34
4         NA          NA       FALSE          NA    12
5      FALSE        TRUE       FALSE       FALSE     3

Here is a mock-up of a barplot I thought I could make to present the counts of cases with the different combinations of attributes:

I'm probably not searching with the right keywords but I just can't quite find any examples for the type of plot I want to make, so I'm not sure how to do it. Do I need to combine a heatmap with a barplot or is there a better way of achieving this? What is a plot like this actually called (if there is a name)?

CodePudding user response：

Not exactly what you desire, but might be a start. edit: Dirty hack to het the colnames in an empty bar

library(tidyverse)

df %>% 
  rbind(c(names(df)[1:4],0)) %>%
  unite("label", starts_with("att"), sep = "\n") %>%
  mutate(count = as.numeric(count)) %>%
  ggplot(aes(x = label, y = count))  
  geom_col()  
  labs(x = "")

CodePudding user response：

If you want to do it all in a single ggplot call, without stitching plots together, you could do:

library(tidyverse)

ystep <- max(df$count, na.rm = TRUE) / 5

df %>% mutate(A = paste(attribute_A), B = paste(attribute_B),
              C = paste(attribute_C), D = paste(attribute_D)) %>%
  ggplot(aes(x = 1:5, y = count))  
  geom_col(fill = 'deepskyblue3', color = 'deepskyblue4', width = 0.5)  
  scale_y_continuous(limits = c(-max(df$count), max(df$count)),
                     breaks = c(4:1 * -ystep, pretty(df$count)),
                     labels = c(names(df[4:1]),  pretty(df$count)))  
  geom_tile(aes(y = -1 * ystep, fill = A, color = A), height = ystep)  
  geom_tile(aes(y = -2 * ystep, fill = B, color = B), height = ystep)  
  geom_tile(aes(y = -3 * ystep, fill = C, color = C), height = ystep)   
  geom_tile(aes(y = -4 * ystep, fill = D, color = D), height = ystep)  
  geom_text(aes(y = -1 * ystep, label = A, color = A), fontface = 2)  
  geom_text(aes(y = -2 * ystep, label = B, color = B), fontface = 2)  
  geom_text(aes(y = -3 * ystep, label = C, color = C), fontface = 2)  
  geom_text(aes(y = -4 * ystep, label = D, color = D), fontface = 2)  
  scale_fill_manual(values = c('pink', 'gold2', 'lightgreen'))  
  scale_color_manual(values = c('red4', 'orange4', 'green4'))  
  theme_minimal()  
  theme(panel.grid.major.x = element_blank(),
        panel.grid.minor.x = element_blank(),
        axis.text.x = element_blank(),
        axis.title.x = element_blank(),
        axis.title.y = element_text(hjust = 0.75),
        legend.position = 'none')