Bayes' Theorem — the Test Paradox

What just happened?

Even a very accurate test (99% sensitivity and specificity) for a rare disease produces plenty of false alarms. The reason: for a rare disease the number of healthy people is huge, so a small percentage of false positives among the healthy outweighs the true positives among the few sick. That is why a positive result often means a low probability of disease.

Bayes’ theorem:

— describes how to update the probability of a hypothesis $A$ in light of new data $B$. In diagnostics: $A$ = "disease", $B$ = "positive result".

Numeric example (per 100,000 people):

At a prevalence of $0.1\%$: $100$ sick → the test detects $99$; among $99\,900$ healthy people, $1\%$ false positives is $999$ people. Out of $1098$ positives only $99$ are truly sick, so $P(\text{d}\mid+)=\tfrac{99}{1098}\approx 9\%$. The probability depends also on the base rate, not only on the test accuracy.

Why does this work? (the invariant)

Ignoring the base rate (prevalence) is the classic "base-rate fallacy" — common even among doctors. The probability of disease given a positive result depends also on how common the disease is in the population, not solely on the sensitivity and specificity of the test. For a rare disease the false positives dominate; the more common the disease, the larger the share of truly sick among all positives (the gauge $P(\text{d}\mid+)$ rises).

History and applications

The theorem was formulated by Thomas Bayes, and his work was published posthumously in 1763. It is the foundation of Bayesian statistics, medical diagnostics, spam filters, machine learning, and rational inference. It teaches us not to trust "test accuracy" blindly without asking: how common is the thing being tested for?

Why you need this for the exam:

This extends conditional probability. Trick: write out concrete counts over a large population (e.g. 100,000) and count how many are true and false positives — then you compute $P(A\mid B)$ directly from the definition, with no base-rate fallacy.

Remember: a positive result from an accurate test for a rare disease usually means you are healthy anyway — because false alarms predominate. Always ask about the base rate before trusting "test accuracy".