Calculating confidence intervals using bootstrapping

Table of contents

1. Overview
2. The technique
3. Implementation

---

Overview

When calculating summary statistics from a sample of measurements, it is often helpful to also calculate confidence intervals for the statistics. Bootstrapping can be a helpful technique for doing so.

The technique

A very helpful description of the technique is provided here - see Section 6 for an example of calculating the confidence interval of the mean of an underlying distribution, using a sample of measurements.

Implementation

Matlab includes a bootci function which can be used to calculate confidence intervals using bootstrapping. I have used Matlab to calculate confidence intervals using bootstrapping when calculating receiver operating curve (ROC) statistics using the perfcurve function.