Risk numbers lab.
One comparison.
More than one way to see it.
Change two invented numbers. See the difference in people, percentage points and relative terms, side by side.
Try the examples
Two fictional groups. The same one-year follow-up.
Imagine 1,000 people in each group. Count each person once if they experience an invented, unwanted event during that year. These are learning examples, not findings from a study or estimates of anyone’s personal risk.
people with the event per 1,000
compared with group A
0.5 percentage points lower in group B.
A relative percentage cannot be calculated when the reference risk is zero. The absolute difference is still available.
Full scale: both bars run from 0 to 1,000 people. The pale space represents people without the event.
| After one year | Group A | Group B |
|---|---|---|
| With the event | 20 | 15 |
| Without the event | 980 | 985 |
| Total people | 1,000 | 1,000 |
| Absolute risk | 2% | 1.5% |
In this invented example, 20 of 1,000 people in A and 15 of 1,000 in B have the event over one year: 5 fewer per 1,000 in B.
Show the calculation
- Signed risk difference: B minus A
- (15 − 20) ÷ 1,000 × 100 = −0.5 percentage points.
- Relative change: difference divided by A
- (15 − 20) ÷ 20 × 100 = −25%.
- Risk ratio: B divided by A
- 15 ÷ 20 = 0.75.
Negative differences mean fewer events in B; positive differences mean more. Relative percentages and risk ratios are rounded for display. Method source
Hold the headline steady. Change the starting risk.
Switch between 20 → 15 and 200 → 150. Both show a 25% reduction, but the difference is 5 versus 50 people per 1,000. Write down both numbers before judging the claim.
Then try 20 → 25. The same arithmetic now describes an increase. Finally, enter zero in group A: the lab keeps the absolute comparison and explains why the relative one is undefined.
Read the complete risk guideArithmetic is one part of reading a study.
Real estimates need uncertainty, a clearly defined outcome and a relevant comparison. These invented counts have no confidence intervals and establish no treatment effect. A hazard ratio, an odds ratio or a result across several studies needs its own interpretation.
The one-year label belongs to this exercise. Do not multiply a one-year percentage to predict a five-year result. Check the actual study’s follow-up and analysis.
Connect numbers to the wider evidenceRead the difference correctly.
Why is 0.5 percentage points different from 25%?
In the starting example, subtract 1.5% from 2%: the gap is 0.5 percentage points. Divide that gap by the starting 2%: it is one quarter, or 25%. The statements describe the same invented comparison. Interpretation source
Does “no difference” mean two treatments are equivalent?
No. The 20-versus-20 button only sets two numbers equal. It contains no trial, uncertainty estimate or equivalence margin. Explore the guide to non-inferiority trials for the questions a real comparison must answer.
Why can I zoom the chart?
The full view keeps all 1,000 people in sight. Zoom makes small event counts easier to compare, but the axis then ends below 1,000. The label states that endpoint, and the table always keeps the full denominator. Zoom changes the picture, never the numbers.
Can I use this as a personal risk calculator?
No. It has no information about a person, condition or treatment. Use it to practise reading invented frequencies. For definitions, see absolute risk, relative risk and confidence intervals.
How this exercise was prepared.
Prepared with AI. AI checked the relevant passages in the public HTML sources below on 27 September 2026. No human editorial or clinical review has been completed. The numbers, exercises and interface are original learning examples; they do not come from a clinical trial.
- Cochrane Handbook, chapter 6: effect measures
Public HTML, sections 6.1.2, 6.2.4 and 6.4.1: risk differences, ratios, zero reference risks and follow-up. Relevant passages checked, not a review of every cited paper.
- Cochrane Handbook, chapter 15: interpreting results
Public HTML, sections 15.4.1 and 15.4.3: absolute and relative presentations and frequencies per 1,000.
- National Cancer Institute: absolute risk
Public dictionary entry checked for the concept and its time frame. No cancer-risk estimate from the entry is used here.
- National Cancer Institute: relative risk
Public dictionary entry checked for the comparison between groups.