Aggregates, privacy and chart critique
A good critique checks the unit, calculation, scale, identity and claim using actual numbers. Published aggregates can reveal individuals when filters overlap.
An aggregate can still reveal a person (W6 p. 30). A table reports four students in a group, three passed: individual results are unknown. A second filter shows three of those students, all passed. Subtracting, failures in the full group = 4-3 = 1 and in the filtered group = 3-3 = 0, so the one excluded student must be the failure. Overlapping published filters allow such differencing. Protections include minimum cell sizes, suppressing or merging tiny groups, and checking what combinations of releases reveal.
Design task (W6 p. 31). Constructed inspections: line A inspected 200 items with 20 defective; line B inspected 50 with 10 defective.
- More items to rework? Plot counts: A has 20, B has 10. A has the larger rework volume.
- Higher defect rate? A: 20/200×100 = 10%; B: 10/50×100 = 20%. B has the higher rate, 10 percentage points (or twice) A's.
A supported claim: "In these inspections, line A produced more defective items (20 versus 10), while line B had the higher observed defect rate (10/50 = 20% versus 20/200 = 10%)." It does not say why B's rate is higher.
Chart review checklist (W6 p. 32):
| Check | Question | Answer for the example |
|---|---|---|
| Unit | What does one record mean? | One inspected item |
| Calculation | Where does 20% come from? | Line B: 10/50 × 100 |
| Scale | Where do the bars start? | 0, on a clearly labelled common axis |
| Identity | How can readers identify groups? | Production-line names label the bars |
| Claim | Does this prove a causal effect? | No; observed comparison only |
A useful critique points to actual numbers, scales and labels in the figure, rather than saying "the chart is confusing".
Source: W6 PDF pp. 1-37. Most numbers in the lecture are illustrative or constructed for teaching; pp. 26-27 reuse the Week 5 synthetic registration counts. The Week 6 lab manual (L6) practises Week 5 EDA reasoning and is worked through section by section in Lab work.
Practice problems
Supplementary study questions on the lecture concepts — not verbatim exam questions. Each answer counts toward spaced review.
1. Conceptual understanding
Definitions and distinctions.
Line A: 20 defects in 200 items. Line B: 10 in 50. Which has the higher defect rate?
Four students: three passed. A filter showing three of them has three passes. What is revealed?
2. Application
Apply the concept to a case or calculation.
Which chart answers which line has more items to rework?
In the factory example, what does one record mean?
3. Interpretation and limitations
Evaluate assumptions and justify conclusions.
What helps prevent differencing disclosures?
Which is a useful critique?
Explain it in writing
Self-checked, not marked. Write a few sentences, then compare with the example answer. Your writing is saved as you type.
Show an example answer
The full group has 4 − 3 = 1 failure and the filtered group has 0, so the one excluded student failed. Two harmless-looking aggregates have revealed a person's result.
Check: did you answer the question, use evidence or working, and avoid claiming more than the evidence supports?
Show an example answer
“In these inspections, line A produced more defective items (20 vs 10), while line B had the higher observed defect rate (10/50 = 20% vs 20/200 = 10%).” It describes the comparison without claiming a cause.
Check: did you answer the question, use evidence or working, and avoid claiming more than the evidence supports?