Data Visualization: The Convergence of Scientific Data and Value-based Graphics
Written by Justin Sodano, PA, Head of Global Studio (US) on Thursday 11th April 2024
It requires a multidisciplinary and multi-departmental approach, not only for collecting and interpreting the data, but also for creating effective data visualization outputs. Leveraging the skills, techniques, and software programs from data scientists, outcomes researchers, and graphic designers, in combination with insights from Medical Affairs representatives (e.g. medical directors, medical scientific liaisons [MSLs], etc.), can help optimize the meaning and effectiveness of an information graphic.
The principles
As highlighted in our recent sessions, Edward Tufte, the "da Vinci of data" or "Galileo of graphics”2 outlines six principles on how to portray data accurately. These include:3
- Comparisons: Illustrate the similarities or differences between two or more dependent variables
- Causality: Depict how a change in one or more independent variables influences a dependent variable
- Multivariate: Compile and transform varying sets of complex data into an easily interpretable narrative
- Integration: Combine information from multiple sources into one coherent dataset
- Documentation: Include appropriate attributions, titles, and measurements
- Context: Provide information that is essential for understanding the data and incorporate trendlines, when appropriate, to predict future results
How to “Data Viz”
Once you have assembled your data visualization “go-to” team, identified the statistical and graphic design programs that are at your disposal within your organization, and understand the principles of data visualization, you can start to conceptualize your information graphic.
To do this, ask yourself the following questions outlined in our simple 1,2,3 approach, based on IBM’s data visualization best practices4:
- Who is your audience?
- What is the best visual(s) to use?
- How can it be simplified or refined?
Who (audience)
Before even putting pen to paper or pushing pixels, knowing your audience is key. The type of visual or chart type you choose hinges on the audience with whom you plan to communicate. What type of information is most important to that audience? Does it need to be a basic graph with aggregate data for a generalist or patient (i.e. written in plain language)? Or does it need to be a more elaborate visual with patient-specific data demonstrating the finer points of individual responses for a specialist? Knowing the level of knowledge of your target audience, the types of information they are interested in, and what might be most impactful to them will help guide your decision on what data to include or exclude.
What (chart type)
The type of chart(s) you choose is fundamental in data visualization. Understanding what chart types are available within the programs you are using and what they best represent are essential when creating your visual. Do your data set and key point lend themselves to a comparison of categories, which is best illustrated with a column graph, horizontal bar graph, or line chart? Is there a need to demonstrate a relative relationship between two or three variables, which can be achieved with a scatter plot or bubble graph? Are your data more like a static composition or “parts-to-a-whole” story, which may be best presented as a donut graph, tree map, starburst graph, stacked bar graph, or circle packing graph? Does your composition graph include an element of time, such that a stacked bar graph, stacked area graph, or Sankey diagram would be appropriate? Or do your data focus more on the distribution of data over time, which could be represented by a scatter plot, line histogram, or beeswarm plot?
How (simplify/refine)
The beauty of data visualization is pure, accurate simplicity. The ability to take complex multivariate data and present them in a simplified, easily accessible manner is what makes great data visualization so powerful. In the end, that is what data visualization is intended to do: communicate key information quickly and accurately, while leaving a memorable impression on the viewer.
How does one simplify graphs? By removing the unnecessary chart clutter. There are many terms for non-essential graphical elements or data. Fundamentally, you must focus on what information adds value to your data visualization story and what is distracting or misleading. Is it necessary to have an axis if all the bars have data labels? Is there a need for horizontal grid lines that may make some viewers go cross-eyed? Think about the amount of time your brain needs to process each element and color on a graph. Less is more – the fewer unnecessary elements or less information your graphic has, the more impact it will have on the audience.