Bootstrap Distribution of a Statistic
You'll learn: How to generate thousands of bootstrap resamples to build an empirical distribution of any statistic from your data.
What Is a Bootstrap Distribution?
After you've learned how bootstrap resampling works (drawing samples with replacement from your original data), the next step is to repeat that process many times and collect the results. Each resample gives you one value of your statistic (like a mean, median, or standard deviation). When you do this thousands of times, those values form a bootstrap distribution—an empirical picture of how that statistic varies due to sampling.
Think of it like this: Imagine you're trying to estimate the average height in a room by measuring 10 people. If you could remix those 10 people thousands of times (putting names back in a hat after each draw), you'd get thousands of slightly different averages. The spread and shape of those averages show you the variability in your estimate—that's your bootstrap distribution.
Why This Matters
The bootstrap distribution gives you:
- Standard error (the spread of your statistic)
- Confidence intervals (the range where the true parameter likely falls)
- Shape information (is it symmetric? skewed?)
All without needing mathematical formulas or assuming your data is normal!
The Process
- Take a bootstrap resample (sample with replacement, same size as original)
- Calculate your statistic (mean, median, etc.)
- Store that value
- Repeat steps 1-3 many times (typically 1,000–10,000)
- You now have a distribution of that statistic
Key Takeaway: The bootstrap distribution is built by generating thousands of resamples and computing your statistic each time—it shows you how much that statistic naturally varies and forms the foundation for bootstrap confidence intervals.