INFOVIS

information visualization

interactive visual representations of data to amplify cognition. Purpose = insight, not pictures

two main purposes of InfoVis

Analysis and presentation

analysis

figuring out questions to ask, when you don’t know what you’re looking for

presentation

Communicating ideas, visual storytelling, data journalism

representation

Chart types (bar, line, etc.)

interaction

Zoom, hover, scroll, filter

Simpson’s Paradox

Trends in sub-groups reverse when aggregated; Always visualize data to uncover hidden patterns

Nominal data attribute

Categories, no order (e.g., house vs condo)

Ordinal data attribute

Ordered categories (e.g., hot, warm, cold)

interval data attribute

Continuous numeric (e.g., age, % exam score)

position, size, orientation, color

visual encoding channels by precision

Expressiveness, Effectiveness, Redundant Coding

3 key encoding principles

Expressiveness

Visual encoding should express all of, and only, the information in the

dataset

effectiveness

Importance of the attribute should match the salience of the channel

redundant coding

Using multiple visual encodings to represent one variable

Why are visualizations powerful but limited

They offload work to vision, but perception has systematic illusions & biases

Ebbinghaus Illusion

Surrounding areas influence judgment of size

Muller-lyer illusion

Arrow directions affect perceived line length

sine illusion

Viewers perceive difference between two lines as changing when it’s actu-

ally constant

change blindness

Failure to notice large changes in a visual scene

pre-attentive processing

bottom up, Fast (<250ms), automatic, stimulus-driven, e.g., finding a red circle among blue

attentive processing

(top-down): Slow, sequential, goal-driven, requires working memory, e.g., finding 3s in a grid

Quick comparisons, highlight outliers, Can mislead, bias patterns

Pro and Cons of pre-attentive processing

figure/ground, proximity, similarity, symmetry, closure, common fate, connectedness, continuity

8 gestalt principles

figure/ground

Distinguishing foreground from background (ex: fedex)

proximity

stronger cue than similarity, elements close together are perceived as a group

similarity

similar elements perceived as belonging together (rows dominate)

symmetry

Symmetrical elements seen as belonging together

closure

We complete incomplete shapes mentally

common fate

Elements moving in same direction are grouped (ex: animations)

connectedness

Connected elements perceived as a group, Can override similarity and proximity

continuity

Prefer continuous lines to abrupt changes

anomolies, cluster, correlate, characterize distributions, compute derived values, extremes, range, filter, retrieve value, sort/rank,

Amar, Eagan & Stasko (2005) 10 tasks

correlation coefficient

Range –1 to +1. Small=0.15, Medium=0.30, Large=0.50. Correlation ≠ causation

analyze, search, query

Munzner’s (2014) high-level actions

Small multiples, Parallel coordinates, Parallel sets, Star plots, Heat maps, Mosaic plots

how to handle 3+ variables?

parallel coordinates

Encode variables along horizontal axes; data = polylines.

• Positive correlation → lines don’t cross

• Negative correlation → lines cross

guideline for color usage

Avoid rainbow maps, use single-hue sequential scales for continuous data, diverging color for midpoints, be aware of label effects

Clear, precise, efficient communication, more ink for data

Graphical Excellence (Tufte)

Distortion from scale, 3D, and size, ideally 1

Lie Factor

Declutter, Focus, Consistency, Aesthetics matter

4 design guidelines

Accuracy, Precision, Reliability

3 aspects of uncertainty

Error bars lead to within-the-bar bias, gradient plots, violin plots

Why avoid error bars? Alternatives?

Frequencies (“1 in 4”), unit charts, animation

Best way to communicate risk

Magazine style, Annotated chart, Infographic, Slideshow, Film/Video, Data comics, Flow chart

genres of data stories

Author-driven vs. Reader-driven

Author-driven = fixed order, heavy text, limited interactivity. Reader-driven = flexible, open, interactive

Martini glass, Interactive slideshow, Drill-down story

3 story structures?

Highlighting, Animation, Ordering, Annotation, Interactivity, Metaphors

Visual narrative tactics?

unit charts

Part-to-whole, probability, humanizing data, when symbols recognizable

affordances

Properties of an object that determine how it could be used; design must match task

Information, List, Timeline, Comparison, Map, Statistics, Flowchart, Hierarchy, Anatomical, Animated

types of infographics