Understand the idea
An overall result may hide different error rates in different groups. Evaluation needs context, adequate data and attention to consequences.
WORKED EXAMPLE
In a fictional classifier test, Group A has 1 error in 10 cases and Group B has 4 in 10. Equal sample sizes do not make the error rates equal.
Watch it step by step
Step 1 of 4Count
Record A: 1/10 errors and B: 4/10.
Choose Play to advance steps automatically.
Read or choose any step
- Count
Record A: 1/10 errors and B: 4/10.
- Calculate
A error rate is 10%; B is 40%.
- Compare
The difference is 30 percentage points.
- Investigate
Ask why and seek more evidence before deployment.
TRY BEFORE CHECKING
Exercises with feedback
MAKE SOMETHING YOU CAN EXPLAIN
Your practical challenge
Write an evaluation note for the fictional results. Include group rates, sample-size limits and a question to investigate.
These ticks are your self-review. They do not automatically assess project quality. Ask an adult or partner to review what you can show.
Use the skill in a new situation
Explain how false positives and false negatives could have different effects.