Quiz dotyczący ponownego próbkowania Bootstrap

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. Which sampling procedure follows the ordinary bootstrap rule for the original observations 2, 4 and 9?
   • Draw two observations without replacement from the three originals
   • Draw three observations with replacement, allowing a sample such as 2, 2 and 9
   • Draw four observations with replacement from the three originals
   • Draw all three observations without replacement on every repetition
2. Which mean belongs to the illustrative bootstrap sample 2, 2 and 9?
   • 5
   • 3
   • 4
   • 13 divided by three
3. When comparing two models on identical test cases, what must remain together in each resampled unit?
   • The case outcome and both models' predictions
   • Only the two models' overall scores
   • A prediction from one case and an outcome from another
   • All positive outcomes without their predictions
4. Which endpoints define the simple two-sided 95% percentile bootstrap interval described?
   • The 2.5th and 97.5th percentiles of the original observations
   • The 5th and 95th percentiles of the bootstrap statistic distribution
   • The 2.5th and 97.5th percentiles of the bootstrap statistic distribution
   • The minimum and maximum of the bootstrap statistic distribution
5. What does increasing the number of bootstrap resamples primarily reduce?
   • Selection bias in the original sample
   • Simulation noise in the resampling calculation
   • Dependence between the original observations
   • Uncertainty caused by an entirely missing population group
6. Several records come from each participant. Which choice better reflects that grouping when appropriate to the study design?
   • Resample every record independently without considering participants
   • Resample each participant's individual rows separately and treat them as independent people
   • Keep every participant fixed and randomly reassign their records across participants
   • Resample participants together with their associated records
7. Bootstrapping test cases for one fixed fitted model does not automatically account for which uncertainty?
   • Variation from retraining and selecting the model
   • Variation in the sampled evaluation cases
   • Variation in the recomputed metric across resamples
   • Simulation variation caused by finite repetitions
8. Why can a narrow bootstrap interval still be misleading when the original dataset excludes an important population group?
   • Every bootstrap interval is invalid for model metrics
   • A narrow interval proves the missing group behaves identically
   • Resampling cannot supply evidence about conditions absent from the sample
   • The percentile method automatically estimates that group's outcomes