Art by Jake Upton
By Kayla Downs
Graphs and other figures are critical to understanding complex scientific concepts, trends, and data. They are one of the most important tools in science yet they are often incorrect, misleading, or hard to understand. There is an art to effective data visualization and it can completely change the way something is understood. Making data clear and easy to visualize is crucial to a widespread understanding of science.
Visual learning is a primary mode of understanding but technical scientists often lack the design knowledge or software skills that are needed to create the best visuals for their data. This is a skill that can be acquired with some basic principles but it also highlights why the partnership between scientists, designers, and artists is so important. Stephen Midway, Ph.D, is a professor and researcher of fisheries ecology at Louisiana State University (Bachelor’s from UVM!). His article ‘Principles of Effective Data Visualization’ outlines the ten principles to make data visualization as appealing as possible that he’s found through his data-heavy research.
#1: Diagram First
Don’t think of the technical specs here, think about what you are trying to convey. Get a clear vision of your purpose and message and what data you want to share. Do this before you pick up a pencil or open a software program!
#2: Use the Right Software
There are many software programs out there that can create graphs and visuals. Some just have the basic tools while others can be more complex. Learning to use these requires practice but having a strong command of them will improve your work immensely.
#3: Use an Effective Geometry and Show Data
The shapes and figures are what is most important for an appealing graphic. Most geometries fall into the categories of amounts (comparisons), compositions (proportions), distributions, and relationships. Use the best option for your data and include your data in the graph!
#4: Colors Always Mean Something
Colors convey meanings and studies have shown that colorful graphs are more memorable. It is best to use highly contrasting colors, such as red and blue as well as colors that can be converted into grayscale without losing the meaning.
#5: Include Uncertainty
Leaving out uncertainty can lead to misleading visuals and an incomplete statistical representation. One of the clearest ways to show uncertainty is to include the data points on a graph.
#6: Panel, when Possible (small multiples)
Paneling means to repeat a figure to show differences. Because the design elements are the same, it is easier to see the differences in data.
#7: Data and Models are Different Things
While raw and summarized data are often straightforward, a model may require more explanation. Decide if it is effective or not to include a model.
#8: Simple Visuals, detailed Captions
Not only should captions fully explain the geometry used in the model, they need to fully explain the visualization and representation used. Ideally, the model and caption should be able to be taken out of the context of the paper and still be completely understood.
#9: Consider an Infographic
Infographics typically include more text, images, and ideas than a single data model. They were also shown to have the highest memorability. While not all models should be converted to infographics, many models can benefit from some annotations.
#10: Get an Opinion
Seek external opinions on your model before it gets published. The best visuals are the ones that readers connect with.
It is clear that there is an art to designing visuals that are effective at conveying data, visually appealing to look at, and make statistics easier to understand. These principles serve as guidelines, but there is no formula for making the best visual. This process requires researchers to tap into their creative design skills and be the artist for their data!
Midway, S. R. (2020). Principles of Effective Data Visualization. ScienceDirect, 1(9). Retrieved February 22, 2021, from https://www.sciencedirect.com/science/article/pii/S2666389920301896.