What are some effective techniques for visualizing uncertainty in data?

2 How to choose a technique?

When choosing a technique to visualize uncertainty, there are several factors to consider, such as the type and level of uncertainty you want to show. Is it in a single variable, a relationship, a distribution, a prediction, or a comparison? Additionally, the format and scale of your data is important. Is it categorical, numerical, temporal, spatial, or multidimensional? Is it continuous, discrete, or ordinal? The goal and context of your visualization should also be taken into account. Do you want to explore, explain, or persuade with your visualization? Who is your audience and what is their background and interest? How will your visualization be displayed and accessed? Moreover, the number of data points should be taken into consideration.

3 What are some examples of techniques?

Visualizing uncertainty in data can be done in several ways, each with their own advantages and disadvantages. Error bars and confidence intervals are a simple and common way to indicate the uncertainty in the mean or median of a variable, or the difference between two groups. However, they may not be accurate if the data is not normally distributed or if the error bars are too wide or too narrow. Shaded areas and ribbons are similar to error bars, but they show the uncertainty in the entire range or distribution of a variable. They can be confusing when the shaded areas overlap or obscure the data points. Violin plots and bean plots are alternatives to box plots that reveal the uncertainty in the shape and density of the distribution of a variable, but they may not be familiar or easy to understand for some audiences. Error bands and fan charts are used to show uncertainty in trends or forecasts over time, though they can be difficult to interpret. Heat maps and contour plots are ways to show uncertainty in spatial or multidimensional data sets, but they may not display outliers or individual data points. Hypothetical outcome plots and ensemble plots are used to show uncertainty in simulations or scenario analyses, though they may appear noisy, cluttered, or overwhelming.

4 How to apply a technique?

Once you have chosen a technique for visualizing uncertainty, you should follow some best practices to apply it effectively. To start, use appropriate and consistent scales, axes, and labels for your data and uncertainty measures. For instance, use logarithmic scales for skewed data, percentage scales for probabilities, and clear labels for your variables and units. Additionally, consider using visual cues such as color, size, shape or texture to encode your uncertainty. For example, lighter or darker colors can indicate lower or higher uncertainty respectively. Moreover, interactive and dynamic features like tooltips, sliders, filters or animations can be used to enhance your visualization. Furthermore, explanatory and narrative elements like titles, captions, annotations or stories should be used to communicate your visualization. For instance, titles can summarize the main message or insight of your visualization while stories can guide your audience through the logic of your visualization.

5 What are some challenges and limitations?

Visualizing uncertainty can be difficult and has its limitations. For instance, some data sets may not have error estimates, some uncertainty measures may not have standard definitions or methods, and some uncertainty comparisons may not have valid or meaningful baselines or benchmarks. Additionally, some uncertainty concepts may not match the common sense or expectations of your audience, some uncertainty visualizations may not convey the intended message or impression of your data, and some uncertainty decisions may not have clear or optimal criteria or outcomes. Moreover, some uncertainty situations may not be appropriate or relevant for your purpose or context, some uncertainty disclosures may not be welcome or beneficial for your audience or stakeholders, and some uncertainty manipulations may not be honest or responsible for your data or impact.

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?