Goal
Reinforce 2 lessons, add a third
Visualization for a purpose
- the more you know about the purpose, the better you can to achieve it
- not knowing is a purpose
Purpose makes it safe to throw stuff away: less purpose, throw less away
special purpose may be OK – since you can have multiple views
Mappings and encodings
math sense and catography sense of maps
- many possibilities
- explore!
Think Differently! (Task Bravery)
obvious mappings/encodings aren’t the only ones
- unusual rotations / projections
- distortions, non-linear, …
costs and benefits to novelty
Simplification is only one of the ways we get to make choices
what is excess to one person is essential to another
TASK (or user) CENTRICmay not be one-size-fits all
"Bad" Visualizations can change framing
Hand-Drawn things as inspiration
different sets of constraints, easy to try different things
Over-Simplified Model of Vis
better model when we get to nested model
Domain Task/Objects –> Data –> Variables –> Visual Variables
Abstraction
can shoehorn most stuff into this.
Choose encodings: visual variables to data variables
Sampling
(left over from last time)
Use 1D events (analog to snow)
overdraw, binning (histograms), kernel density estimates, pareto chart, rotation (look at spaces), other designs
in hindsight – you can know what’s right
Encodings and Mappings
Visual Variables:
- Position, color, shape, intensity, texture, orientation, …
Data Variables:
- Need to abstract to have different types (coming next week)
- Categorical vs. Ordered vs. Quantitative
- Relative vs. Absolute
- Metric vs. Non-Metric
- Local Comparison vs. Non-Local Comparison
John Snow’s Map
Death (event), position, time ==> x,y, mark
Position is the most prominent visual variable
Tie it to something important
Use it as a secondary thing (to allow for placement)
no direct meaning, but put points in relationships
Distorting Maps (task bravery)
- Cartography
- Fisheye Views
- Image Retargeting
- Metric Spaces
- Mathematics of Mappings (what is preserved)
Route Maps
Piles of technical details you don’t need to worry about (the algorithms)
The Distortions:
- Length Distortion
- Angle Distortion
- Shape Distortion
Each throws something away: but what do you get for it?
Tradeoffs: they even say “performed carefully”
Shape simplification that preserves key features
I would be interested in learning why this projection hasn’t been implemented over the past 10 years, as it seems to be relatively sound.
It was – the company went out of business
We question their goals: (make something really good at X)
How did they do at achieving their goals?
Some of their new things do the opposite! (different goals)
Short Route Maps
Not a practical technique (doesn’t scale)
Interesting because it takes the space warp idea to the limit
Clear why it doesn’t work
But when does it work? And what does that tell us?
Are the warp-based route maps fundamentally flawed?
(probably impactical – glad someone figured it out)
negative result?
What can we learn from them?