Bayes' theorem & base rates
About this app
A mosaic of prevalence × sensitivity/specificity: why a positive test for a rare disease is often no reason to panic. The false-positive area beats the hit area — posterior P(D∣+) live, the base-rate fallacy made visible.
The statistics
Bayes’ theorem reverses a conditional probability: from a test’s sensitivity P(+|D) it gives P(D|+), the probability of actually having the disease after a positive result. P(D|+) = P(+|D)·P(D) / [P(+|D)·P(D) + P(+|¬D)·P(¬D)]. The key ingredient is the base rate P(D), the prevalence. For a rare disease there are so many healthy people that even a small false-positive rate produces more false positives than true positives. Ignoring this is the base rate fallacy.
What you see
A unit square represents the whole population: on the left the sick group, as wide as the prevalence, on the right the healthy group. Each column is split by test result, giving four fields: true positive, false positive, false negative and true negative. A second chart shows the step from prior to posterior. A values panel lists prior, sensitivity, specificity, P(D|+) and P(D|−) and the four counts per 10,000 people, with Bayes’ formula alongside.
Things to try
Load the rare disease example: with 1 % prevalence, 99 % sensitivity and 95 % specificity the false-positive field is larger than the true-positive field, and P(D|+) lies far below the sensitivity. Then push the prevalence up and watch the posterior rise sharply. Raise the specificity to 99.9 % and compare which slider has more effect. The common disease example shows the situation when the condition is widespread.
Frequently asked questions
- What does Bayes’ theorem say?
- Bayes’ theorem links the conditional probabilities P(A|B) and P(B|A): P(A|B) = P(B|A)·P(A)/P(B). It describes how a starting probability (prior) is updated by new information into a posterior. In the medical example it turns a test’s accuracy into the probability of being ill after a positive result.
- What is the base rate fallacy?
- The base rate fallacy is neglecting how common a feature is in the population. A typical case is equating a test’s sensitivity with the probability of being ill after a positive result. For rare diseases that probability is often much lower, because many positive results come from healthy people.
- What is the difference between sensitivity and specificity?
- Sensitivity P(+|D) is the share of sick people the test correctly flags as positive. Specificity P(−|¬D) is the share of healthy people it correctly classifies as negative. One minus specificity is the false-positive rate.
- What can I set in the Bayes app?
- You set prevalence, sensitivity and specificity with sliders or load the rare and common disease examples. The mosaic, the step from prior to posterior, P(D|+), P(D|−) and the counts per 10,000 people update instantly.
Subject: Statistics
More from Statistics
- Anscombe's quartet
- Bias-variance / overfitting
- Box plot workshop
- Geometric mean
- Law of large numbers
- Histogram workshop
- Hypothesis test & p-value
- Confidence interval
A VisuApp by heyprof: interactive, free in your browser, no sign-up. What is a VisuApp? · All apps