CDT502 Chap.5 Visual Evidence and Data Stories
Visual Evidence and Data Stories
This chapter selects charts from the analytical comparison rather than a software gallery. It ranks position and length ahead of angle, area and colour for precise comparisons, connects distributions to histograms, boxes and dots, and uses scatterplots with residuals for relationships. Titles, scales, denominators, annotations and narrative verbs are audited so that design clarifies evidence without strengthening the claim.
Begin by writing the comparison a reader must make. Values aligned on a common position scale support close quantitative judgement. Length also works when a meaningful baseline is preserved. Angle, area and colour are weaker for small numerical differences but useful for showing grouping, emphasis or broad pattern. The visual channel should reduce the cognitive work required by the analytical task.
Chart choice follows variable roles. A numerical distribution can be shown through dots, bins, density or quartiles, each preserving different detail. Group comparisons require aligned scales and visible sample context. Two numerical variables call for a scatterplot, followed by residual and influence checks if a line becomes the claim.
Ordered time belongs on an ordered axis; connecting unordered categories with lines invents continuity. Design parameters remain analytical choices. Histogram width, smoothing, axis limits, category order, faceting and colour palettes can change what the audience notices. Record settings, keep identity colours stable and provide another cue when colour distinguishes a critical group.
An annotation should explain why evidence matters rather than repeat a coordinate. Narrative provides sequence without licensing stronger conclusions. Lead with the decision, establish the baseline and denominator, present the evidence, then state uncertainty and action. Titles must use verbs supported by the design.
Before publishing, ask whether an omitted group, time period or denominator could reverse the message, and regenerate every figure from the same code that produced the reported values.
What this chapter covers
- 01
Visual attention and channels
- 02
Chart choice by variable role
- 03
Distribution displays
- 04
Relationships and residuals
- 05
Colour, facets and scales
- 06
Narrative and publication checks
Worked example · free
Test whether one branch creates the fitted slope
- 2Fit and inspect the relationship with every branch.
- 3Run a sensitivity fit and inspect residual patterns.
- 3State the association and causal boundary.
Key terms
- Visual Channel
- A property such as position, length, area or colour used to encode data.
- Small Multiples
- Repeated charts using consistent encodings to compare groups with less overplotting.
- Permutation Importance
- The held-out performance loss produced by disrupting one feature's information under a fitted predictive model.
Visual Evidence and Data Stories FAQ
How do I choose between a histogram and a boxplot?
Use a histogram when distribution shape, peaks or gaps matter and record the bin rule. Use a boxplot for compact comparisons of median, middle spread and unusual values across groups. When the dataset is small or individual values matter, add a dot plot rather than asking either summary to show everything.
Why can a chart title be statistically wrong?
A title is a claim. Words such as drives, improves or reduces imply causation that an observational comparison may not identify. Write the observed contrast, period and groups first. Reserve causal verbs for a design that controls assignment or otherwise establishes the counterfactual needed for that claim.
What should I check before publishing a chart?
Check the unit, denominator, period, missing-value rule, baseline, scale, category order, colour meaning, direct labels and uncertainty. Confirm that the intended comparison uses an accurate channel and that omitted groups or periods would not reverse the message. Regenerate the chart from the analysis code.
Exam move
Write the comparison as a sentence before opening chart software. Map the quantities being compared to position on a common scale where possible. Practise showing the same observations with dots, boxes and histograms, explaining what each reveals and hides. Review residuals whenever a line becomes the central story.
Redesign one chart each week by changing only a single choice: baseline, category order, bin width, palette, faceting or annotation. Explain how that change alters the reading task without changing the data. Build a title ladder from purely descriptive wording through association to causation and stop at the verb the design supports.
Recreate a figure from code after deleting the exported image so reproducibility becomes routine. Ask another reader what comparison they noticed first; if it differs from the intended one, revise the encoding rather than adding a longer caption.
Working through Visual Evidence and Data Stories in CDT502? Sia is AskSia’s AI Statistics tutor — ask any CDT502 Visual Evidence and Data Stories question and get a clear, step-by-step explanation grounded in how CDT502 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
