Rhetorical Elements

Gestalt Theory Principles

Preattentive Processing Features

Guidelines for Better Visualizations

Form and Function (can be plotted on a grid)

Qualitative Data

non-numerical information (usually text) collected through direct observation, interaction, or archival research
iconography can be potent by is culturally situated and audience dependent
word clouds and trees trade detail and sophistication for engagement, but they can be nuanced with categorical small multiples
sentiment analysis tracks the emotional valences of terms through compiled dictionaries, but those connections are subjective, contingent, and may not measure what they purport
n-grams (phrases of n length) can be very useful to identify themes within a corpus

Comparison Charts

Purpose drives choices of data representation
Don't misrepresent data by doing things such as breaking a bar graph or using a non-zero axis
Use less ink so greater data density is possible
Conventional formats such as bar graphs are often more legible but less engaging
Icons can be used to make proportions comprehensible

Time Charts

Most time chats are versions of a line or area chart
- the timeline is a master trope in our engagement with visual information
Charts can build on familiar metaphors
- such as the clock face and a circular line chart
Time charts can have different ethical conventions than comparison charts
- you may not need to start the axis at zero (although we must still avoid deception)
Purpose drives our choices of representation
Charts can bear more data aspects than they conventionally do
- for example, varying the line width (height, thickness, or stroke weight) of a Gantt chart allows it to track an additional variable

Distribution Charts

Distribution can address forms of uncertainty
- uncertainty from randomness: statistical confidence in models
- uncertainty from unknowns: incomplete or poor data
Showing the data can be effective through strip plots, beeswarm plots, hexagonal binning, Wilkinson dot plots and wheat plots, raincloud plots, or stem-and-leaf plots, for example
Visualization choices as always are driven by purpose and audience
- We're not "dumbing things down"; we're conscious of our audience and goals

Relationship Charts

Correlation is not causation but they are definitely associated
Chart types can be used for a variety of purposes; we tend to associated tree diagrams with biological taxonomy and organizational charts, but they can plot a range of relationships (particularly if you consider horizontal structures)
All charts offer trade-offs suited to purpose and audience
Arc charts can show relationships but can be distorted by proximity
The datasaurus underscores the importance of understanding the information and its representation; the same numbers can be presented in very different, even contrary, ways