Special Considerations for Data Visualization
By Jen Christiansen / 3 minute read
Data visualization can come in many forms, from simple charts to complicated online interactives. While time, skill, and story scope will determine where on that spectrum you land, the fundamentals are the same.
On the science beat, much of the data you’ll be working with have been analyzed and peer-reviewed. That doesn’t mean it’s flawless, though, so you must still retain a critical eye. Read the “methods” sections of papers and data portals. Become familiar with the questions that guided data collection, as well as the subjects of the study — particularly relevant in regard to health stories, as the results may be rooted in a homogeneous sampling of people, a detail that should inform how you frame the data.
Spot-check data provided by scientists against their published figures, keeping an eye out for red flags. Are outliers simply dropped from final figures without note in the published paper? Ask the corresponding author (that is, the scientist flagged on that paper as the point person for queries) for an explanation. Having a hard time replicating straightforward figures from a published paper? Ask the corresponding author about the methods as a reality check before moving forward with other visualization solutions.
I often find it useful to pinpoint which figure in a scientific paper presents the crux of the findings.
Increasingly, data are being made available for direct download in tandem with published research articles. The quality of the data available in open access formats can vary. Sometimes they’re the raw data, sometimes they’re already processed. Ask the corresponding author for guidance if anything is unclear. When dealing with highly processed and specialized data — genetics and astronomy pop to mind — I find it useful to work with freelance data visualizers who are domain experts.
Nadja Popovich, a graphics editor on The New York Times‘s climate team, notes that “one of the best parts of working on science-centric graphics is that scientists and other researchers by and large want to talk to you. They are often more than happy to help you better understand the data, so that you can represent it correctly.” Her best advice for science-centric graphics is her best advice for all data graphics, “When in doubt, talk to an expert.”
Overwhelmed by the possibilities, and faced with a fast-approaching deadline? I often find it useful to pinpoint which figure in a scientific paper presents the crux of the findings. Then ask the researchers for the data they used to build that specific graphic, and think through how to present the same data in a format that would make sense to your nonspecialist audience. Look at figures in the supplemental information repository for this purpose. You can often find gems hidden there. Such was the case with one of my favorite visualizations, by the designer Jan Willem Tulp, for the story “Trillions of Insects Migrate,” which was inspired by reviewing a figure hidden in a paper’s supplementary information.
As with all data visualizations, be cognizant of data-collection methodologies. This is especially true with global health data, in which the methods are rarely consistent over space and time. Seek advice from content experts when trying to determine which authoritative source is considered the best option for the subject and/or regional focus of your story. Upon plotting out the data in your preliminary explorations, question any and all surprising patterns. Does the prevalence of a disease rapidly change? Check the documentation to make sure that the shift isn’t due to a change in data-collection methods, and then check with a content expert before jumping to conclusions. Sometimes the right decision is to not visualize the data at all. Or, as Amanda Makulec writes in “Ten Considerations Before You Create Another Chart About COVID-19,” “do more to understand the numbers.”
Katie Peek, an independent data-visualization designer and science journalist, notes that the following principles — which she learned during her time as a scientist — continue to serve her well as a graphics editor:
- If something seems too good to be true, it often is. Cast a careful eye on any surprising results, and make sure they’re right.
- Check your work. Do the calculation another way, or find another data set or expert, to confirm that your result is reasonable.
- Just because something is published doesn’t mean it’s right. A peer-reviewed journal article can still be wrong.
- Keep your analysis approachable. If you can’t easily explain what you’ve done to the data to present them in the graphic, you’re probably doing too much.