Lecture: kartičky | information-visualization

best channel to convey information = vision

pre-attentive perception

the time to find the blue dot is constant – no matter the total number of dots
- assuming we can distinguish the colors

visual information-seeking mantra

visual perception is a two stage process

Gestalt psychology

one stimulus, two perceptions
there is a difference between stimulus and perception
emergence – we need to see the whole picture, not just parts
reification – perception contains more spatial information than the stimulus
multistability – ambiguous stimuli can generate different perceptions but they cannot coexist simultaneously
invariance – objects are recognized independently of various variations (transformations, lightning, …)
laws of grouping – law level perceptions are grouped into higher-level objects
- good Gestalt

information visualization pipeline

source data → data tables
- data transformations
data tables → visual abstraction
- visual mappings
- transition from data form to visual form
- we use data attributes to create (visual) marks
visual abstraction → views
- view transformations
- result … actual pixels

example: gapminder (data attributes → visual channels)

interactions

the user can manipulate 1) data transformations, 2) visual mappings, 3) view transformations
zoom … view transformation
filtering … data transformation
in gapminder, we can also edit visual mappings

taxonomies of data types

“what comparisons can I make?”
“how can I aggregate the data?”
nominal … only equality (=)
- aggregation … mode (we show the most frequent one) or top-k
  - if we have a taxonomy (hierarchy), we can group the data
ordered … ordering and equality (<, =)
- aggregation … median, quantiles, histogram (count per bucket)
quantitative … “how much smaller is it?”
- → intervals … $v-v'$
- → ratios … $v/v'$ $v / v^{'}$
  - only if there is a meaningful zero on the scale
- aggregation … mean, std dev, skew