Bayes' Theorem Example: Surprising False Positives

Understanding Test Accuracy: False Positives and False Negatives
📌 Medical tests can yield false positives (test says positive, but the condition is absent) or false negatives (test says negative, but the condition is present).
🧪 A scenario with a 5% false positive rate illustrates that a positive result does not guarantee the condition, especially if the disease is rare.
📉 Initial intuition suggests a 95% chance of having the disease if the false positive rate is 5%, but the actual probability is significantly lower, dependent on the disease's rarity.

Application of Bayes' Theorem
🤔 Bayes' theorem is used to formally calculate conditional probabilities, relating $P(A|B)$ (probability of A given B) to $P(B|A)$ (probability of B given A).
🔢 In the medical context, A is having the disease, and B is testing positive; the formula combines the prior probability of the disease, the probability of a positive test given the disease (related to false negative rate), and the probability of a false positive.
💡 With a 1% disease prevalence, a 10% false negative rate (meaning 90% true positive rate), and a 5% false positive rate, the probability of actually having the disease after one positive test is approximately 15.4%, much lower than the intuitive 95%.

Iterative Updates with More Information
🔁 Performing a second positive test significantly updates the belief; the probability of two consecutive positive results given the disease becomes 0.90 0.90 = 81%.
📈 After two positive tests, the probability of having the disease rises from 15.4% to approximately 77%.
🔑 The rate at which an event occurs in the general population (prevalence) is crucial; new information, like symptoms or demographic data, allows for Bayesian inference to continuously update and refine probability assessments.

Key Points & Insights
➡️ The rarity of the underlying condition dramatically influences the reliability of a single positive test result.
➡️ A single positive result when disease prevalence is only 1 in 100 only yields about a 15.4% chance of actually having the condition, contrary to common assumptions.
➡️ Bayesian analysis demonstrates that accumulating new information (e.g., a second test or knowledge of high-risk demographic) allows for continuous, objective updating of probabilities.
➡️ Always be clear about underlying assumptions and use new evidence to adjust the calculated probabilities accordingly.

📸 Video summarized with SummaryTube.com on Mar 04, 2026, 09:20 UTC