Data Visualization
85 flashcards by MentisLibrary
Chart choice, visual encodings, color, scales, and accessibility, grounded in Tufte and Cleveland's research on how people actually read graphics. Includes the classic ways charts mislead.
- What is the purpose of a data visualization?
- To let a reader see a pattern, comparison, or relationship faster and more reliably than from a table
- How do you choose a chart type?
- Start from the task: comparison, distribution, composition, relationship, or change over time
- When should you use a bar chart?
- For comparing a value across discrete categories
- Why must a bar chart's axis start at zero?
- Because bar length encodes the value, so truncating the axis distorts the ratio the reader sees
- When is a line chart the right choice?
- For continuous change over time or another ordered variable
- What is a scatter plot for?
- Showing the relationship between two quantitative variables
- What does a histogram show?
- The distribution of a single quantitative variable across bins
- What does a box plot show?
- The five-number summary and outliers, which makes it good for comparing distributions across groups
- What is a violin plot?
- A box plot combined with a density curve, showing the full distribution shape
- What is a heatmap best used for?
- Showing a value across two categorical or binned dimensions using color
- What is a small multiple?
- A grid of small charts with identical axes and encoding, one per category
- Why are pie charts often criticized?
- Because angle and area are judged less accurately than length or position
- When is a pie chart acceptable?
- For two or three parts of a clearly defined whole where only rough proportions matter
- What is a stacked bar chart good and bad at?
- Good at showing totals and the largest segment, bad at comparing middle segments across bars
- What is a 100 percent stacked bar for?
- Comparing composition across categories when the totals do not matter
- What is a slope graph?
- Two points per item connected by a line, showing change between two periods and reordering
- What is a waterfall chart for?
- Showing how a starting value reaches an ending value through sequential positive and negative contributions
- What is a dot plot and when does it beat a bar chart?
- A point per value on a category axis. It beats bars when the values are far from zero or the list is long
- What is a bubble chart?
- A scatter plot where a third variable is encoded by point area
- What is a treemap for?
- Showing hierarchical part-to-whole composition using nested rectangle areas
- What is a sparkline?
- A small, word-sized line chart embedded in text or a table to show a trend inline
- What is the Cleveland and McGill ranking of visual encodings?
- Position is judged most accurately, then length, angle, area, and finally color intensity
- What are preattentive attributes?
- Visual properties the eye detects almost instantly, such as color, size, orientation, and position
- What is the data-ink ratio?
- The proportion of ink that encodes actual data rather than decoration
- What is chartjunk?
- Decorative elements that add no information and can obscure the data
- What is Tufte's lie factor?
- The ratio of the size of an effect shown in the graphic to the size of the effect in the data
- Why is a truncated y-axis dangerous?
- Because it magnifies small differences and can reverse the reader's impression
- When is a logarithmic scale appropriate?
- When the data spans orders of magnitude or when proportional change is the subject
- What is wrong with dual y-axes?
- The relative scaling is arbitrary, so any apparent correlation can be manufactured
- What is banking to 45 degrees?
- Choosing an aspect ratio so the average line slope is near 45 degrees, which makes slope changes easiest to judge