How to show confidence intervals in Datawrapper bar charts
Using a confidence interval to show how certain you are (or your data source is) about averages, or adding target markers, are great ways to show more nuance in your visualization. You can add both to Datawrapper bar charts with the “overlays” option.
This article explains how to add confidence interval/error bars and value markers to Datawrapper bar charts like in the example above. If you've never created a Datawrapper bar chart before, we recommend reading How to create a bar chart first.
How to show confidence intervals in bar charts
Upload the data
First, our chart needs some data. To create a bar chart with a confidence interval, you need to upload four columns:
- One with the categories. These will show up as the labels in front of the bars.
- One with the bar values. These will be visualized as the length of the different bars.
- One with the lower bound of the confidence interval, and...
- ...one with the upper bound of the confidence interval. These two columns mark the minimum and maximum value of your overlay.
|White, not Hispanic
|Hispanic (any race)
Upload your data in step 1: Upload data, then proceed to step 3: Visualize.
Prepare the bar chart
Once you selected bar chart as your visualization type, continue to the Refine tab. Here, make sure that the column with your bar values is selected in the Select column dropdown:Now you can customize your bar chart to your liking; e.g. by changing the colors, label alignment, etc. You can find more information about this in our article Customizing your bar chart.
Add the confidence interval overlay
To add the overlay that will work as our confidence interval, scroll all the way down in the Refine tab until you see the button Add overlay. Click on it and some few new options will open. Here, select Range.
If you want to see just one marker, you can also choose Value instead.
These are the options you have for the Range overlays:
- Column lets you select your two columns that you want to use as the lower and upper bound for the range overlay.
- Title let's you create a color key-like item above the chart that explains to your readers what exactly they see. Since there are different kinds of confidence interval (90%, 95%) and value overlays can be used not just for error bars, we recommend to make use of the Title.
- With Color, Opacity and Pattern, you can change the appearance of the value overlay. Depending on how important you is, you might want to decrease or increase the opacity. If your style guide prefers opaque colors, you can use the striped patterns to still create a look-through effect.
In a 2014 study, Michael Correll and Michael Gleicher (PDF) found that people judge values inside the bar to be more likely than values outside the bar (when a classical error bar is used). This is called "within-the-bar-bias".
After adding an overlay, ask yourself: Which values would you judge as more likely when seeing this visualization? Can you change the opacity, color, or pattern of the overlay to work against within-the-bar bias?
You could also make the bar completely transparent (e.g. with the hex code
#ffffff00) to just show the ranges. This way, they look similarly important on both sides:
If you have any questions about creating confidence intervals in bar charts, don't hesitate to get in touch with us at firstname.lastname@example.org.