A hands-on guide for applied researchers

What would your estimate look like, if you'd sampled again?

The nonparametric bootstrap answers that question without a second data collection wave. It treats your one sample as a stand-in for the population, draws new samples from it with replacement, and recomputes your statistic each time — building up a picture of how much that statistic would bounce around if the world handed you a different sample.

STEP 01

Bring in a sample

Type your own values, paste a list from a spreadsheet, or generate a sample from a known distribution to see the logic play out on data you already trust.

Values (comma, space, or newline separated)

Shape

Sample size (n)

Random seed (optional — same seed, same sample)

Your sample, plotted at true value (each point is one observation)

STEP 02

Draw resamples, with replacement

Each resample below is drawn from your n observations — same size, drawn with replacement, so some observations appear more than once and others not at all. We'll draw this many resamples, and compute one statistic from each.

Statistic to track

Mean Median Std. deviation 10% trimmed mean

Number of resamples (B): 1000

B = 100 B = 1000 B = 5000

Resamples will appear here as they're drawn — the first several are shown in full so you can see the mechanics; the rest are drawn instantly and folded into the distribution below.

STEP 03

Read the bootstrap distribution

This histogram is the distribution of your chosen statistic across every resample — the bootstrap's estimate of how that statistic would vary from sample to sample.

bootstrap resamples statistic in your original sample 95% bootstrap interval

Peek under the hood — see the first resamples' actual draws