Take a disease one person in a thousand has, and a test that flags 99% of the sick and 1% of the healthy. The chance that someone picked up by the test is actually sick is 0.99 × 0.001 in the numerator, over 0.99 × 0.001 + 0.01 × 0.999 in the denominator, and the workbench answers 0.09016393442623. A 99% accurate test, and nine out of ten flagged people are wrong.
The two terms explain it if you look at their sizes: true positives arrive at 9.9⋅10−4 per person, false positives at 0.00999 — ten times the stream. The same arithmetic in natural frequencies is easier on the intuition: screen a thousand people, one is ill and is caught with probability 0.99, while 9.99 of the 999 healthy ones are flagged anyway. Nothing about the test is inaccurate; what is scarce is the disease, and a 1% error rate applied to a healthy population swamps a 99% hit rate applied to a sick one.
Answer
0.09016393442623
How to read it
The positive predictive value on the workbench. The two following links are its numerator and the false-positive term, which is the larger of the two.
Source
Positive predictive value = sensitivity·prevalence / (sensitivity·prevalence + false-positive-rate·(1 − prevalence)); base-rate neglect was demonstrated in Kahneman and Tversky's taxi-cab problem and the screening version popularised by Gerd Gigerenzer
Work it in the calculator
These links open the right calculator with this problem's numbers already typed in. The line above is what the screen should show — check it yourself.