Tomas Petricek
email: t.petricek@kent.ac.uk
twitter: @tomaspetricek
office: S129A
Open data collected by governments, e.g. data.gov.uk
Not so open data collected by governments
Social networks and other user data
Data collected by businesses or scientists
Multimedia data such as photos, videos, books
Devices, sensors and internet of things
The goal is to aid our understanding of data by
leveraging the human visual system's ability to
see patterns, spot trends and identify outliers.
Scientific data visualization
Helping experts make sense of complex data
Telling stories with data in media
Make data and facts more accessible
Understanding user needs
Using user experience methods
Do people care about understanding data?
I think that the people of this country have had enough of experts...
Can good visualization get people interested in data?
Principled mapping of data variables to visual
features such as position, shape, size and color.
What are the variables?
What visual features?
What tasks does it enable?
Could perception mislead us?
Form follows function
Beyond conveying facts
Types of data attributes
Types of quantitative scales
More types are maps, hierarchies and networks
Cannot be measured and ordered
Categories represented as shapes
Categories represented as colors
Can compare and guess distance
X position
GDP (quantitative)
Y position
Age (quantitative)
Bubble size
Pop. (quantitative)
Bubble color
Cont. (categorical)
X position
Country (category)
Y position
CO2 (quantitative)
Bar color / offset Year (category)
Can encode categories and magnitudes
How to use it correctly
Choose
colors that colorblind people can distinguish!
Use luminance scale that is perceived as linear!
Rainbow can model magnitude (but not linear by default) or categories.
Position
(+/- 2x)
Length
(+/- 3x)
Angles
(+/- 5x)
Areas
(+/- 6x)
Data analytical task (high-level)
Discover new hypothesis
Present some discovery
Nature of search (mid-level)
Lookup (know where) or locate (know what)
Browse general area of interest
Querying of data points (low-level)
Identify information about data point
Compare multiple data points
Poor choice of visual channels
Misusing our ability to spot patterns
Ignoring implicit channel properties
Misleading perception of scaling
Area is harder to see than position
Easy to confuse length (radius) and area
Humans are too good at seeing patterns!
Implied correlation with too few data points.
Value is position on common axis, not length!
Position channel suggests zero as minimum.
Value is length!
(height of the barrel)
Perception of three-dimensional objects is misleading and difficult.
Allow exploration of large data
Tell a story through data
Element of surprise in visualization
Make the viewer think critically
Visualization for data exploration
Visualization for data presentation
Animation makes the point stronger.
Neat use of radial projection for circular value!
Scrollytelling
Adapts standard online reading interaction.
You draw it!
Can data visualization make you think more critically?
What is data visualization?
Principled mapping of data variables to visual
features such as position, shape, size and color.
Data variables
Visual channels
Remember error rates of visual channels!
What you should remember from this lecture
Tomas Petricek
t.petricek@kent.ac.uk | @tomaspetricek
Related courses
Textbooks and papers
Online links