Course contentsShow
← Data Science
Lesson 301 of 2,1458. Confidence IntervalsFree lesson

Percentile Bootstrap Confidence Intervals

Construct confidence intervals by taking percentiles directly from the bootstrap distribution of your statistic.

Percentile Bootstrap Confidence Intervals

You'll learn how to construct confidence intervals by simply reading values directly from your bootstrap distribution.

The Core Idea

Remember how you created a bootstrap distribution by resampling your data thousands of times and calculating your statistic each time? Now comes the payoff: to build a confidence interval, you just find the percentiles that match your desired confidence level.

For a 95% confidence interval, you want the middle 95% of your bootstrap distribution. That means:

  • Find the 2.5th percentile (the lower bound)
  • Find the 97.5th percentile (the upper bound)

Think of it like trimming the extreme 2.5% off each tail of your distribution. What remains in the middle is your confidence interval.

Why This Works

Your bootstrap distribution approximates how your statistic would vary if you repeated your study many times. The middle 95% of that distribution represents the range where your statistic "usually" falls. This gives you a data-driven confidence interval without assuming your data follows any particular distribution (like normality).

Example in Plain Terms

Suppose you bootstrapped the median income 10,000 times and got 10,000 median values. Sort them from smallest to largest. The 250th value (2.5% of 10,000) is your lower bound, and the 9,750th value (97.5% of 10,000) is your upper bound. Done! That's your 95% confidence interval for the median.

For a 90% confidence interval, you'd use the 5th and 95th percentiles instead (trimming 5% from each tail).

Key Takeaway: Percentile bootstrap confidence intervals are constructed by taking the appropriate percentiles directly from your bootstrap distribution—no formulas, no normality assumptions, just sorted resampling results.