Conditional Probability and Bayes Theorem

The one idea: A test result and the truth it's testing for are two different events, and "how accurate the test is" (P(positive | disease)) is not the same number as "how much you should trust a positive result" (P(disease | positive)). Bayes' theorem is the formula that converts the first number into the second — and when the thing being tested for is rare, the gap between them can be enormous.

⚠️ The trap: "The test is 99% accurate, so a positive result means I'm 99% likely to have the disease." That feels right and it's wrong. A 99%-accurate test applied to a rare disease produces mostly false positives, not mostly true ones — build the tree from the actual base rate before you believe a single test result, because accuracy describes the test, not your odds given a positive.

A screening test comes back positive for a disease that affects 1 in 1,000 people. The test correctly flags 99% of people who actually have the disease, and correctly clears 99% of people who don't — by any everyday reading, a genuinely excellent test. What's the probability this particular person actually has the disease?

Most people, including most physicians surveyed in the classic studies this scenario is drawn from, answer somewhere around 95–99%. We're going to work the actual arithmetic below, and the true answer is under 10%. Nothing about the test is broken. The gap between the intuitive answer and the real one is exactly what this lesson is about: conditional probability, and the theorem — Bayes' theorem — built to handle it correctly.

Conditional probability: the base ingredient