Inspection Paradox, explained.
The inspection paradox occurs when observing a process at a random time disproportionately samples its longer intervals, making your experience differ from the average interval.
Why it happens
Long gaps occupy more of the timeline, so they are easier to land inside. For random arrivals between renewal events, waiting depends on the variability of intervals as well as their mean. Averaging the timetable alone misses that weighting.
Sampling a process at a random time weights intervals by their length.
Read the result
Change the share of long gaps and compare the ordinary mean interval, length-biased interval, and random-arrival wait. These summaries weight the same gaps differently; changing their share changes both the mean and the variability.
A worked example
Two bus timetables with the same mean
A perfectly regular bus every ten minutes gives a random passenger an average five-minute wait.
If each gap is independently five or fifteen minutes with equal probability, the mean gap remains ten, but the average random-arrival wait is 6.25 minutes.
Long gaps catch more passengers. Equal mean service intervals need not imply equal waiting experiences.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
Short intervals last 2 minutes and long intervals last 10. The mean interval is the frequency-weighted average; a random arrival sees the length-biased average E[L²]/E[L]. Its expected residual wait is E[L²]/(2E[L]) for uniformly sampled time inside intervals.
Where this idea is useful
A practical use
A passenger arriving without a timetable experiences a different average bus interval than the average printed gap.
A common misconception
“Half the average gap is always the average wait.”
That shortcut works for regular intervals. Variability increases the chance of arriving during an unusually long interval.
What this explanation leaves out
- This is an ideal stationary renewal process with two interval lengths; scheduled arrivals change the result.
Where else does length-biased sampling appear?
A randomly encountered ongoing job may be unusually long, and a random user may experience a busier period. The sampling method determines what receives extra weight.
Are you sampling events equally, or sampling the time those events occupy?
Further reading
Explore the original research or the teaching reference behind this experiment.