Navigate
Courses Tutorials Practice Pricing Tools Upgrade
Check your understanding

Quiz

Across seven lessons you turned a single guess into an honest account of what you do not know. You separated a prediction interval (about one new outcome) from a confidence interval (about the average), then built coverage you could actually certify: split conformal from held-out misses, conformalized quantile bands and prediction sets that grow when the model hesitates. You fit conditional quantiles with the pinball loss, put a model's probabilities on trial with calibration, and got uncertainty from resampling alone with the bootstrap and the jackknife+. Running through all of it is one discipline: a number is only honest if it comes with a range you can defend, matched to the decision it feeds. This quiz checks what stuck. The last two steps are live R you can run.

Question 1 of 8

Which range do you quote?

A client is selling one specific 1,500 square foot flat and asks what it will fetch. You hold a confidence interval of $323k to $335k and a prediction interval of $250k to $408k. Which do you quote, and why?

Right. A confidence interval is about the average line and shrinks toward it with more data; a prediction interval adds the irreducible spread of an individual home, which is the only thing that can keep a coverage promise about this sale.
Averaging two intervals that answer different questions answers neither. The client asked about one home, so the prediction interval is the correct object, wide as it is.