MAST20034 · Critical Thinking With Data
Honest Graphics and Visualisation
Week 2 turns the critical lens onto pictures. A graph is an argument, and a good one shows the data, not the decoration — so this chapter codifies the five graphics principles (show the data, encode honestly, maximise the data-to-ink ratio, label clearly, choose the right form) that the exam uses as a marking rubric. You learn to choose the chart type by purpose (comparison, distribution, relationship, composition, trend) and to read the variable type off the prompt before picking a form. The signature exam skill is the graph-critique skeleton — the markers literally reward the pattern two good features + one specific fix — and you drill the classic misleading-graphics tricks (a truncated or dual axis, area distortion, cherry-picked scales) by name. Finally you read distributions properly: the boxplot as a five-number summary and the histogram shape, including the mean–median gap that reveals skew. This is a high-frequency exam chapter: a figure to critique appears on nearly every paper.
What this chapter covers
- 012.1 Exploratory data analysis (EDA) — look before you analyse
- 022.2 Variable type drives every chart choice
- 032.3 The graph-critique skeleton (two good features + one specific fix)
- 042.4 Misleading graphics — truncated/dual axes, area distortion, cherry-picked scales
- 052.5 Boxplots — the five-number summary and outliers
- 062.6 Histograms — shape, skew and the mean–median gap
Critiquing a graph — the two-good-one-fix skeleton, mark by mark
- +1Two good features: the chart uses a clear categorical x-axis (two comparable years) and labels both bars — the comparison it intends is legible.
- +1Name the specific flaw: the vertical axis is truncated (starts at $90m, not $0), so the bar area no longer encodes the value honestly — a 3% rise looks like a near-doubling.
- +1Explain the consequence: truncation breaks the encode-honestly principle; the reader's eye reads bar height as magnitude, so the graphic overstates growth.
- +1State the fix: begin the axis at $0 (or, if a zoomed view is genuinely needed, use a line chart with a clearly marked broken axis and state the true percentage change).
Key terms
- Five graphics principles
- Show the data; encode honestly; maximise the data-to-ink ratio; label clearly; choose the right form. The exam's de facto marking rubric for any ‘critique this graph’ prompt.
- Data-to-ink ratio
- Tufte's idea that most ink in a graphic should encode data, not decoration. Chartjunk — 3D effects, gradients, needless gridlines — lowers the ratio and obscures the message.
- Truncated axis
- An axis that does not start at zero, so bar heights or areas no longer encode magnitude proportionally. The single most common honest-encoding violation; it exaggerates differences.
- Boxplot
- A five-number summary drawn as a box (Q1–median–Q3) with whiskers and plotted outliers. Compact for comparing groups and spotting skew, but it hides multimodality a histogram would reveal.
- Skew (mean–median gap)
- The direction a distribution's tail runs. Right-skew pulls the mean above the median; left-skew pulls it below. The gap between mean and median is a quick read on skew and on which centre to quote.
Honest Graphics and Visualisation FAQ
How do I structure a graph-critique answer?
Follow the skeleton the markers reward: name two genuine good features, then one specific, named fix — citing the graphics principle each point relates to. Specificity (‘truncated y-axis’, not ‘misleading’) is what earns the mark.
When should I use a boxplot versus a histogram?
Use a boxplot to compare several groups compactly or to flag outliers; use a histogram when the shape of a single distribution matters — modes, gaps, skew — because a boxplot can hide bimodality.
What are the most-tested misleading-graphics tricks?
A truncated or dual y-axis, area/volume distortion (scaling both width and height for a one-dimensional value), and cherry-picked scales or time windows. Learn them by name so you can label them instantly in a critique.
Exam move
Memorise the five graphics principles as a rubric and the two-good-features-plus-one-fix skeleton as your answer template — a figure to critique is on nearly every paper. Build a quick mental catalogue of named misleading tricks (truncated axis, dual axis, area distortion) so you can label rather than describe. Practise reading skew off a histogram via the mean–median gap, and reading a boxplot's five-number summary at a glance. Always tie each critique point back to a principle by name; that is what turns an observation into a mark.
