Evidence & Base Rates
Follow true and false flags through a natural-frequency table, then compare the expected counts with a repeatable synthetic batch.
Notice · 1 of 3
Try the experiment, then check the ideas behind it. Stars track these answers; they do not measure your experiment.
Progress stays in this browser only. First-try answers earn 100 points, second tries 75, and later tries 50. One star per completed checkpoint. The timer is optional and never affects points.
Theory and model assumptions
P(D | +) = ps / [ps + (1 − p)(1 − specificity)]. Expected TP = Nps, FP = N(1 − p)(1 − specificity), FN = Np(1 − s), TN = N(1 − p)·specificity. Sample mode draws defect and flag states with a seeded pseudorandom generator. This is a fictional inspection example.
Teacher guidance
Sensitivity is P(flag | defect); the displayed posterior is P(defect | flag). Expected counts may be fractional. Seeded batches fluctuate and are synthetic. A zero denominator has no defined conditional probability. No independent second-test assumption is built into this activity.
Expected time: 10–20 minutes.