• Central tendency – where is the data located
• Mean (μ), median, mode
• Dispersion / variability – how spread out is the data
• Variance, standard deviation (σ), inter-quartile range, coefficient of variation (σ/μ), mean absolute deviation
• Shape – what are the “tails” of your data?
• Skewness / kurtosis, percentiles, min/max
• Frequency – categorical data

• Mean (μ)
• Most common measure, basis of most statistics
• Weighs observations differently, sensitive to outliers
• Median
• The “typical” value, with half above & half below
• Immune to outliers
• Mode
• The “peak” of the distribution, most common value
• Can have multiple modes, but may be less informative

• Standard deviation (σ)
• Most common dispersion measure, basis of most statistics
• Strongly affected by outliers (because math), normality
• Inter-quartile range
• Robust to outliers, good for skewed data, interpretable
• Coefficient of variation
• Dispersion as % of mean, comparable across units
• Mean absolute deviation (or w/ medians)
• Like σ, but less sensitive to outliers

• Skewness – asymmetric “tails”
• Positive / right skewed: mean > median, more high values
• Negative / left skewed: mean < median, more low values
• Some models assume symmetrical data (skewness = 0)
• Kurtosis – heavy or light “tails”
• High  heavy tails, low  light tails
• Percentiles – bins of data
• Further resolution of full distribution, less of a summary

• Count of values
• Most common value
• For ordinal data, 5-number summary
• Min, 25%, 50%, 75%, Max

• Human capacity for processing information is limited
• Human brain adapted to visual pattern recognition
• Improved understanding of complex data
• Identification of patterns, trends, and outliers
• Effective communication of analytical findings to various audiences
• Inform decisions – insight, reasoning, understanding

• Leveraging strengths of human cognition
• Visual bandwidth1: About 20 Mbps (1Gbps total)
• Pre-attentive processing2: Instinctive processing
• Graphical perception3: Process some designs better
• Mitigate constraints of human cognition
• Avoid overloading limited “working memory”4
• Leverage external recognition

1

2

3

4

• Concisely conveying data
• Telling stories for retention
• Relevant XKCD

Relevant XKCD

• A story is a set of data, observations, or events that are presented in a way or order intended to evoke a reaction or specific conclusion
• Can be as simple as well designed visualization
• Can be fully narrative, without tables or figures

• As the global population grows, much of that growth is happening in cities. In 1950, 70% of the global population lived in rural areas. In 2050 that is expected to decline to 30%

• Global populations are flocking to cities
• In 2050, that is expected to fall to 30%

• Start with main idea
• What conclusion do you want your audience to draw?
• Consider audience and format, customize to match
• Find a hook – something to grab attention and anchor memory
• Storyboard – lay out the structure of your story
• Slides, post-it notes, pen and paper. Flexibility is key.

• Types of graphs
• Learning through visualization