Interactive explainer
Interactive: Bootstrap: estimating uncertainty by resampling
The nonparametric bootstrap repeatedly samples the observed data with replacement. Variation in the recalculated statistic estimates sampling uncertainty; it does not create new information about unobserved cases.
Interactive: Bootstrap: estimating uncertainty by resampling
Choose mean or median, the number of synthetic observations and the first case’s duration. Resampling runs locally in your browser.
Reading the display
The curve connects histogram counts from 2,000 bootstrap estimates. It shows the distribution of the estimated statistic, not the original processing times. The text reports the estimate, central percentile interval and standard error. “Values as a table” lists the original observations so that the input remains inspectable.
The random seed is fixed. Restoring a setting reproduces its result. Increasing the observation count generates more synthetic input cases; it does not mean bootstrap has extracted additional real information from the existing cases.
Three experiments
- One long case: raise the first duration from 10 to 60 minutes. Watch the estimate and interval width change.
- Change the statistic: compare mean and median on the same values. Which business question suits each statistic?
- More original cases: compare 30 and 100 observations. Examine how a single large value’s influence and the estimated uncertainty vary.
Reset restores the mean, 30 cases and a ten-minute first case. This simple bootstrap does not model dependencies between cases or unobserved risks. The full explanation gives an example whose resamples can be enumerated completely and discusses the limitations of percentile intervals.