APG5642 Chap.4 Analysing and Visualising Evidence
Analysing and Visualising Evidence
Analysis turns records into a defensible comparison. Before calculating, define the claim in words: what is being compared, over which period, for which population and under which classification rule. This sentence identifies the numerator, denominator and exclusions that must be checked.
Counts answer how many; rates answer how common relative to an appropriate population.
A rising count can coexist with a falling rate when the population grows. Group comparisons require the same definition and time window, while trends require attention to changes in collection practice. Outliers deserve investigation because they may be important events, data errors or evidence that a category is too broad.
Visualisation is an editorial decision.
Choose form by relationship: bars for category comparison, lines for ordered time, dots for precise comparison and scatterplots for relationships between two measures. Start axes and label values in ways that do not exaggerate differences. The chart title should state what is measured, and notes should identify source, period, units and material limitations.
What this chapter covers
- 01
Claims written before calculations
- 02
Counts, rates, shares and appropriate denominators
- 03
Comparable groups, periods and definitions
- 04
Outliers as errors, events or analytical clues
- 05
Uncertainty and limitations in the record
- 06
Chart choice based on the relationship
- 07
Honest scales, direct labels and contextual notes
- 08
Visual evidence integrated with narrative reporting
Recompute a rate before choosing a chart
- 1Calculate North's rate: 84 divided by 12,000 equals 0.007, or 7 complaints per 1,000 users.
- 1Calculate South's rate: 54 divided by 6,000 equals 0.009, or 9 complaints per 1,000 users.
- 1State the correction: North has more complaints by count, but South has the higher complaint rate relative to its user population.
- 1Verify that both districts use the same complaint definition, coverage period and user-population method.
- 1Choose a simple dot or bar comparison of rates per 1,000, with direct labels and a zero baseline for bars.
- 1Add a note that a complaint rate does not by itself measure service quality or explain why complaints were lodged.
Key terms
- Numerator
- The event count placed above the division line when a rate, ratio or share is calculated.
- Denominator
- The relevant population, exposure or total used to place a count in comparable context.
- Comparison group
- A population or period measured with sufficiently consistent definitions to support a meaningful contrast.
- Outlier
- An observation far from the main pattern that may reflect error, an unusual event or a valuable reporting lead.
- Chart baseline
- The reference value from which a visual scale begins, capable of magnifying or muting apparent differences.
- Chart annotation
- A concise label or note that explains a material event, definition, source or limitation within a visual.
Analysing and Visualising Evidence FAQ
When should rates be used instead of counts?
Use rates when groups differ in population, exposure or opportunity and the reporting question concerns relative frequency. Publish the underlying counts too when they help readers understand scale and uncertainty.
How do I know whether two groups are comparable?
Check that definitions, coverage periods, collection systems, inclusion rules and population measures align. If one differs materially, explain the limitation or redesign the comparison rather than presenting a false equivalence.
Which chart type should I choose?
Match the visual to the relationship. Bars or dots compare categories, lines show ordered change over time, scatterplots show relationships and small multiples compare repeated patterns. Choose the simplest form that preserves the evidence.
What makes a visualisation misleading?
Misleading visuals can use inconsistent denominators, truncated bar baselines, uneven time intervals, unexplained exclusions, decorative area that distorts magnitude or labels that imply causation. Honest charts make scale, source and limitations visible.
How should an editor challenge a chart?
Test the strongest alternative denominator, the most influential outlier and any collection change that could affect the pattern. If a reasonable alternative reverses the conclusion, return to reporting, disclose the sensitivity or narrow the claim.
Assessment move
For every exercise, write a claim sentence before opening chart software. Underline the population, period, measure and comparison. Recompute a small sample by hand and inspect outliers before trusting an aggregate. Practise changing raw counts into rates using clearly invented data, then explain when the rate is more informative and what it still cannot prove.
Build a chart-choice notebook with one good and one misleading example for comparison, trend, distribution and relationship. On revision, remove decoration that does not carry evidence and strengthen titles, labels and notes. Run a chart translation drill: describe the visual aloud without using the title, then compare that description with the intended claim.
If the impression is stronger, weaker or different, identify whether scale, ordering, colour or annotation caused the mismatch. Rebuild the chart with direct labels and a transparent note. Finally, write the strongest conclusion the data supports and one tempting conclusion it does not support. This separates presentation skill from evidentiary discipline.
