Course contentsShow
← Data Science
Lesson 302 of 2,1458. Confidence IntervalsFree lesson

Bootstrap Standard Errors

Estimate the standard error of any statistic using the standard deviation of its bootstrap distribution.

Bootstrap Standard Errors

You'll learn: How to estimate the variability of any statistic by measuring the spread of its bootstrap distribution.

The Core Idea

When you create a bootstrap distribution by resampling your data hundreds or thousands of times, you get a collection of possible values for your statistic (like the mean, median, or correlation). The standard deviation of this bootstrap distribution gives you an estimate of the standard error of your original statistic.

Think of it like this: if you repeatedly measured something with slight variations each time, the spread of those measurements tells you how uncertain you are. The bootstrap simulates those "repeated measurements" by resampling.

Why This Matters

Previously, you learned formulas for standard errors (like SE = σ/√n for the mean). But what if you're working with a median, a ratio, or some custom statistic with no neat formula? Bootstrap gives you a universal tool: just resample, calculate your statistic each time, and measure the spread.

The Process

  1. Generate your bootstrap distribution (1,000+ resamples)
  2. Calculate the statistic of interest for each resample
  3. Find the standard deviation of all those bootstrap statistics
  4. That standard deviation IS your bootstrap standard error

Example (Conceptual)

If your bootstrap distribution of sample medians has values like [45, 47, 46, 48, 45, 47, ...] and the standard deviation of those values is 1.2, then your bootstrap standard error for the median is 1.2.

Key Takeaway: The bootstrap standard error is simply the standard deviation of your bootstrap distribution — giving you a measure of uncertainty for any statistic, even when traditional formulas don't exist.