Bootstrap Coverage

Why does each bootstrap sample, on average, asymptotically, include about 63.2 percent of the original data?

Answer

When you draw \(n\) observations with replacement from a dataset of size \(n\), the probability that a specific point is never picked is

\[ \Bigl(1 - \tfrac1n\Bigr)^{n} \;\approx\; e^{-1} \;\approx\; 0.368. \]

Therefore, the probability it is selected at least once is

\[ 1 - \Bigl(1 - \tfrac1n\Bigr)^{n} \;\approx\; 1 - e^{-1} \;\approx\; 0.632, \]
Back to collection