Raising The Bar: Designing Better Bar Charts

Last updated on Dec 18, 2022 3 min read

Introduction

      Bar charts are the workhorse of information graphics. They’re perhaps the most common kind of visualization out there. Bar charts are ideal because they don’t require special training to interpret. They’re simple, clear, and easy to understand. But while bar charts may be simple to consume, they can also be challenging to create.
      This is the first installment in a series that aims to explain some principles – some subtle, some less so – that you should take into consideration as your design your own bar charts, or consume those designed by others. Each installment will cover a different design principle for creating effective and beautiful bar charts. But first: What is a bar chart?

What is a Bar Chart?

      A bar chart (or bar graph) is a graphic that represents numeric data as a set of bars. In technical terms, bar charts compare the level of a quantitative variable across the levels of a categorical variable. In plain English, bar charts visually compare how a quantity varies across a group. In a bar chart, this quantity is represented through the height/length of the bars: the height of each bar represents the magnitude of the underlying value. To illustrate, say the attribute we’re interested in is the total time to perform some task. Then the height of each bar corresponds to the time taken to complete the task. A taller bar means more time. A shorter bar, less.

How can I create a Bar Chart?

GGplot is my preferred data visualization library, so it’s what I’ll use throughout the series. Within ggplot there are two function that you would usually use to create bar charts: geom_bar() and geom_col(). The key difference between the two functions is that geom_bar() performs some computations on the input data before it renders the visualization, whereas geom_col expects the user to perform any summarization beforehand.

data_summary <- apply(iris[, -c(5)], 2 , mean, na.rm = T)
data <- do.call(data.frame, as.list(data_summary))
data <- gather(data, "attribute")

Also for this demo we’ll use the iris dataset that’s built into R. “The famous (Fisher’s or Anderson’s) iris data set gives the measurements in centimeters of the variables sepal length and width and petal length and width, respectively, for 50 flowers from each of 3 species of iris. The species are Iris setosa, versicolor, and virginica.” We’ll simply process the data to compute the average value for each variable: sepal length and width and petal length and width.

For this demo, we’ll use geom_col(). geom_col() expects two aesthetics: x, describes the different groups/bars in the graph; and y, the magnitude(/height/length) for each bar. In our example, the x aesthetic corresponds to the aqttribute column and the y aesthetic corresponds to the value column.

base_plot <- ggplot(data, aes(fill = attribute)) + 
    geom_col(aes(x = attribute, y = value))

base_plot + 
    labs(
        title = "A Basic Bar Chart of 'iris' attributes",
        x = NULL,
        y = "Value (cm)") + 
    theme_minimal() + 
    theme(legend.position = "none")

Conclusion

That’s it for now. In the next installment we’ll start dive into some of the design principle that you should consider when designing or evaluating bar charts. First up: How does the choice of baseline impact a graph?

Data Visualization
Avatar
Akindele Davies
Associate Data Scientist

My research interests include statistics and solving complex problems.

Related