Dataset

The dataset used in this post has 5 columns:

• the first column is called x and actually provides the month of the year, from January to December
• the columns 2 to 5 provide a numeric value for 4 employees of a company respectively. Let’s say it’s the amount of money they brought or spend each month.

The dataset is stored on github and can be loaded as follow:

# Load the dataset that is stored on the web
data <- read.table("https://raw.githubusercontent.com/holtzy/R-graph-gallery/master/DATA/stacked_barplot_negative_values.csv", header=T, sep=",")

This is an overview of how it looks like:

library(knitr)
kable(head(data, 4))

To work with ggplot2 and more generally with the tidyverse, it’s necessary to transform it to a “long” or “tidy” format. This is totally doable thanks to the tidyr package:

# Load the package
library(tidyr)
library(dplyr)

# transform the format
data_long <- gather(data, group, value, groupA:groupD) %>%
  arrange(factor(x, levels = c("Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sept", "Oct"))) %>% 
  mutate(x=factor(x, levels=unique(x)))

# And that's the result!
kable(head(data_long, 4))

Note: sorting the data is annoying and is described more in depth here

Stacked barchart

 # library
library(ggplot2)

# plot
ggplot(data_long, aes(fill=group, y=value, x=x)) + 
  geom_bar(position="stack", stat="identity")

Stacked barchart polish

 # library
library(ggplot2)
library(hrbrthemes)
library(viridis)

# plot
ggplot(data_long, aes(fill=group, y=value, x=x)) + 
  geom_bar(position="stack", stat="identity") + 
  scale_fill_viridis(discrete=TRUE, name="") +
  theme_ipsum() +
  ylab("Money input") + 
  xlab("Month")

Conclusion

By default, ggplot2 uses a “diverging” approach when it comes to stacking negative values. Basically, it places all the negative values below the 0 axis, and all the positive ones above.

If stacking is a topic that interests you, this article goes more in depth.