Frank Elavsky, Data Visualization Specialist
Research Computing Services, Northwestern University IT
1. A Test.
2. Myth or Law?
3. The Basics.
4. Gestalt Principles.
5. The "Problem" in Visualization.
6. A Systematic Approach.
Check your skills.
What does this figure show us?
Source: Lung Cancer Epidemiology
Is this a "good" scientific figure?
Think "effective" instead of "good" - this can be measured and observed.
Great articles by Jen Christiansen of Scientific American:
Source: Steve Haroz
Source: Kellogg Insight
Source: Aaron Geller, FireFly
Chart literacy is important.
Not just for the audience, but for communicators too.
This is the chart version of a chainsaw massacre.
Think about your data! How did this happen?
An honest mistake... or?
Poor designer put their own name on this one.
From Smashing Magazine's series: "Imagine a pie chart stomping on an infographic forever"
An "automated" visualization report.
AKA: the “chainsaws of visual communication”
(creates entry-point and layers)
Hint: it's time to take notes!
Your skill in these will define whether your visualization is successful or not.
The holy triumvirate: Who, why, and where
Who is your visualization for?
This is the first and most important consideration.
I'm not sure even astrophysists know what this means.
Disciplines intersecting over Equation of State. Ozel F, Freire P. 2016. Annu. Rev. Astron. Astrophys. 54:401-40
This color scale is very bad, but if meteorologists did things differently, the public would lose their minds.
Texas Storm Chasers
Why are you visualizing data?
How can you measure your success if you don't define your purpose?
This is explaining data to a general audience - a little fluff is acceptable to keep their interest.
Brain response predicts movie sales.
This is learning data for your own analysis - not meant to be pretty.
Using ggplot to test data
This is exploring the dimensions of the data - meant for the audience to learn something (not you)
"How will Automation Affect Different US Cities?"
Where does the visual end up?
You can do a lot differently depending on the answer to this question.
Editors: "Print. Small size. No color." (Given the context, this isn't too shabby!)
Lung Cancer Epidemiology
Small multiples are great for many static dimensions (if you have the room)
"Public support for vouchers"
How long do you have and how much do you do?
Don't wait until the end of your research or analysis before you consider this.
Tell me why. What is wrong with this?
This is a case of too much dimension for the data.
What can we improve about this? Use empathy. What is hard?
Text is often a big limitation for the context of your visual.
Minimize the time your audience takes to understand your purpose.
The golden rule of visualization.
These gifs courtesy of Dark Horse Analytics
Go to github for the file 'datasets/congress_data.csv' and get to work.
I want you to do better with just 20 minutes more.
You have all leveled up as Chart Wizards. Congrats.
Who is your reader? What about your data will be difficult for them?
Why are you visualizing your data? How will you know when you have succeeded?
Where/what/when are your limitations? Do you have the scope to do what you want?
Frank Elavsky, Research Computing Services, Northwestern IT