Statistics
Bootstrap from One Sample
How much could a statistic vary if this one sample stood in for the population?
Draw one resample from the six observations. Look for a source observation used twice and one left out, then draw 1,000 more.
The observed sample
Six fictional paper gliders flew these distances in meters. A label identifies each flight, even when two distances are equal.
Most recent resample
Draw one resample to see its observations in draw order.
Bootstrap distribution
Each resample contributes one statistic to the distribution.
Histogram counts
| Statistic interval | Resamples |
|---|
Change the sample, statistic, or seed
Teaching notes and definitions
For each bootstrap resample of size n, draw independently n times with replacement from the n original observations. Each observation has chance 1/n at each draw. A new statistic comes from a new resample. Start a new run for a fresh seed. To replay a sequence, select Fix seed before starting the new run. Adding resamples continues the current run.
For the mean, the exact center under this resampling model is the observed sample mean. If σemp is the standard deviation of the original observations using divisor n, the exact standard error of the resampled mean is σemp/√n. The simulation estimates spread with the sample standard deviation of the accumulated statistics. The median averages the middle two values when n is even.
The distribution is conditional on this sample. Additional resamples reduce Monte Carlo variability. They do not supply information about parts of a population that the original sample missed.