4.8/5 - (39 votes)

A medical test that is 99 percent accurate can still be wrong most of the time it comes back positive, and the math behind it has a name: Bayes’ theorem.

It sounds like a contradiction. A highly accurate test producing mostly wrong results. But once the surrounding numbers come into view, the outcome makes perfect sense, and it explains why doctors, code breakers and scientists all lean on the same 18th century idea.

A rule for reasoning backward from evidence to cause

Bayes’ theorem is a mathematical rule for reversing conditional probabilities, letting someone calculate the probability of a cause based on an observed effect. The formula states that the probability of A given B equals the probability of B given A, multiplied by the probability of A, divided by the probability of B.

The theorem links a prior belief about how likely something is, before seeing any new evidence, to a posterior belief, updated once new evidence appears. The posterior probability turns out to be proportional to the prior probability multiplied by how strongly the new evidence supports it, a relationship formalized through what is called the Bayes factor.

Why a 99 percent accurate test still misleads

Imagine a drug test that is 99 percent accurate in both directions, but only 0.3 percent of the population actually uses the drug being tested for. A positive result in that scenario only indicates roughly a 23 percent real probability that the person actually uses the drug, meaning most positive results turn out to be false positives.

A similar medical example makes the point even more starkly. If a certain cancer affects just one person in 100,000, but ten healthy people out of every 99,999 show the same symptoms anyway, the probability that a symptomatic person actually has the cancer works out to only about 9.1 percent. Rare conditions and imperfect tests combine in ways that intuition consistently gets wrong.

See also  A throwaway number from 1925 became a rule scientists live by

A theorem published after its author had already died

Thomas Bayes, an English minister and mathematician, developed the original probabilistic reasoning behind the theorem, but he never saw it published. After his death, his friend Richard Price spent two full years editing Bayes’ unpublished manuscript before presenting it to the Royal Society on December 23, 1763.

Because of Price’s crucial role in rescuing, editing and publishing the work, some statisticians argue the result should really be called the Bayes-Price rule rather than crediting Bayes alone. Pierre Simon Laplace rediscovered the same result independently in 1774, apparently without any knowledge of Bayes’ earlier work, and went on to expand it significantly in his 1812 book Theorie Analytique des Probabilites.

From an 18th century sermon writer to cracking Nazi codes

The theorem’s reach extends far beyond statistics classrooms. In 1941, Alan Turing applied Bayesian reasoning directly to cryptanalysis, using it to help break enemy codes during the Second World War. In 1973, the statistician Harold Jeffreys placed the theorem on a fully axiomatic footing, going so far as to compare its importance directly to the Pythagorean theorem.

There is even a lingering historical mystery attached to it. The statistician Stephen Stigler has suggested that Nicholas Saunderson, a blind English mathematician, may have arrived at the same idea before Bayes did, a claim that remains disputed among historians of mathematics to this day.

A theorem drafted by a minister, rescued by a friend after his death, and later used to help win a world war now quietly powers everything from cancer screening to spam filters, proof that an idea from 1763 can still shape decisions made every single day.

See also  Berkeley seemed to favor men in 1973, until the data was split