Overview
Bayes' theorem updates a probability when you get new evidence. This applies it to the classic test case: how likely is a condition after a positive result?
Enter the prior probability, the test's sensitivity and its false positive rate.
Worked examples
| Inputs | Result |
|---|---|
| Prior Probability of Condition (%)2Sensitivity (%)92False Positive Rate (%)8 | Posterior probability P(ConditionPositive): 19.01%False-alarm share among positives: 80.99% |
Show 2 more examples
| Prior Probability of Condition (%)50Sensitivity (%)80False Positive Rate (%)20 | Posterior probability P(ConditionPositive): 80%False-alarm share among positives: 20% |
| Prior Probability of Condition (%)0.5Sensitivity (%)99False Positive Rate (%)2 | Posterior probability P(ConditionPositive): 19.92%False-alarm share among positives: 80.08% |
How it works
The posterior is the chance of a true positive divided by the chance of any positive. True positives are prior × sensitivity, and false positives are (1 − prior) × false positive rate. The result shows how strongly the prior shapes the answer.
The formula
| P(A) | Prior probability |
| P(B | A) | Sensitivity |
| P(B) | Total chance of a positive |
Practical tips
- A 2% prior, 92% sensitivity and 8% false positive rate give a posterior of 19.01%.
- Use the posterior as the new prior if you test again with an independent test.
- Small changes to a low prior can change the posterior a lot.
Frequently asked questions
Where does the prior come from?
Base rates from research, or a reasonable starting belief before the evidence.
Is this only for medical tests?
No. It applies to spam filters, quality checks and any evidence that updates a belief.
What if the false positive rate is 0?
Then every positive is real and the posterior is 100%.