Lab 04 / Uncertainty
When accurate alerts mislead
Why can a 99% accurate alert be right only half the time?
Change prevalence, sensitivity, and specificity, then inspect every expected true and false alert in a synthetic population. Change one control at a time, then open the values under the chart before interpreting the picture.
View the values behind this chart
Loading the deterministic simulation...
SEEDInterpret the result
The paired guide explains the mechanism, assumptions, and cases where this display should not be generalized.
What this display answers
Accuracy-like inputs do not determine the meaning of an alert by themselves. The starting prevalence controls how many opportunities there are for true alerts and false alerts.
Natural frequencies make the denominator visible: the relevant comparison is true alerts divided by all alerts, not sensitivity alone.
What to notice
- Use the 1% base rate with 99% sensitivity and specificity. True and false alerts are similar in count.
- Raise the base rate while leaving alert performance fixed. The share of true alerts rises.
- Increase the population. Counts grow, but the expected proportions remain nearly the same.
The model behind it
This is a neutral quality-alert model over a synthetic population. Inputs are assumed fixed and every case is assumed to receive the same alert process.
Rounded counts are shown for readability; the theoretical posterior is also calculated before rounding.
Where the result stops
Real systems may have subgroup variation, uncertain parameters, repeated alerts, or changing base rates.
This demonstration is not a medical test calculator, product benchmark, or risk prediction.