Statistics
Decision Rule Simulator
How many late trips should trigger a review?
Predict how lowering the cutoff from six late trips to five will affect false alarms and missed detections. Select Fix seed to compare both rules on the same simulated trips.
Which batches trigger review?
Model probabilities at each cutoff
| Cutoff | Benchmark | Higher rate |
|---|
Risks under the selected rule
Teaching notes and model assumptions
Choose the review cutoff before looking at the observed batch. Let X count late trips out of n independent trips with a fixed late probability p. Then X follows a binomial distribution. If weather or traffic affects several trips together, the independence assumption can fail.
At cutoff k, false-alarm probability is P(X ≥ k) at the benchmark. Power is the same upper tail at the higher rate. Miss probability is P(X < k) at the higher rate. Raising k can only reduce or preserve the chance of a flag at either rate.
The two simulated populations use separate seeded streams. Apply rule draws new batches by default. Select Fix seed and keep the sample size, batch count and rates fixed to compare cutoffs on identical simulated trips. The model probabilities use binomial sums. Simulated proportions vary across seeds.