Evening Rounds Vol. 31
Data Visualization with Anamaria Crisan
At November’s Evening Rounds 31, Anamaria Crisan took us on a journey through the feats and pitfalls of visualizing data. Starting with an exercise, then moving into practical examples, and finally the best processes for figuring out the right way to visualize a given data set, we heard both the overview and the low-down on charts and graphs, numeracy and accuracy, and a practical methodology for making your data shine.
Data Visualization in 3 Questions
So you have your numbers – now what? Anamaria outlined these three essential questions to ask yourself before figuring out how to best represent your data:
-
Why? (Motivation): why do you need to visualize data?
-
What? (Data): What kind of data is being visualized?
-
How? (Visual and Interaction Design): How is data being visualized?
Starting with the Why and What first, and then moving into the How, lets you tackle the purpose of your design before you start designing – a recipe for effective data visualization.
Breaking Down a Visualization
Each data variable you introduce into your visualization deepens the amount of information you’re getting across, but at the same time can make it more difficult to interpret for the person who ends up looking at it.
Variables can be represented with the vertical or horizontal position of shapes, the colour or size, transparency, and more. Anamaria stresses putting your “thinking cap” on to make sure that every variable you’re adding to your visualization is totally necessary – for some wacky (bad) examples, check this site out.
Visualizing Patient Risk
Trying to communicate risk and probability to a patient is sometimes one of the most difficult and pitfall-ridden exercises you can undertake. Anamaria laid out some good, go-to principles for representing data to patients.
First, representing risk or chance as a number or percentage can be difficult to interpret for a lay person (most people are less “numerate” than they are “literate”) – visualization helps people make real sense of data. That being said, all visualizations are not equal!
While every data set is different and requires a different approach, going backwards through the What-Why-How steps can be a great measure of how effective you think your end visualization is
-
How? – Are the visual and interactive design choices appropriate?
-
What? – Are you using the right data, or deriving the right data?
-
Why? – Does the visualizing solve the relevant problem?
After going through your How-What-Why evaluation, Anamaria recommends, if possible, testing out multiple options with physicians or patients to see which ones are most effective and most preferred.
Quick Tips for Designing Your Data
-
Think about your numbers as basic geometric shapes, and build your visualization up from there.
-
Think about how accurate or persuasive you’re trying to be: some graphs, like bar charts, are more accurate and effective than others; some more visually persuasive.
-
While sticking to two dimensions can seem like a constraint, it’s the way most people will digest visual information, and less likely to be unnecessarily complicated.
-
Pen & paper is still a great and under-utilized tool
-
Don’t jump into just visualizing your raw data – explore what you can derive from that raw data
-
Some recommended data visualization tools, from less to more coding knowledge required: Tableau, Shiny, GGVis, RCharts, Processing, D3.JS
-
A great resource Anamaria recommends for anyone trying to visualize health data is www.vizhealth.org – it includes a database of data visualization graphics, each tagged with the type and goal of the visualization, level of detail, and relevant health applications.