1

I came across this limit when considering bootstrapping since $1 - ((x-1)/x)^x$ is the probability that observation $j$ from a sample of size $x$ does appears in a bootstrap sample also of size $x$. In other words, one minus this quantity is the probability of picking $j$ when choosing $x$ elements from $\{ 1, \ldots, x\}$ with replacement.

Using a computer, this limit appears to converge to approximately $0.3678611$ (this is the value for $x=10000$ to 7 d.p). I thought to try applying L'Hôpital's rule, knowing that the derivative of $x^x$ is $x^x(ln(x) + 1)$, but I could not make any progress.

1 Answers1

1

So you have $$ \left(\frac{x-1}{x}\right)^x = \left(\frac{x}{x-1}\right)^{-x} = \left(1+\frac{1}{x-1}\right)^{-x} \to e^{-1} $$

gt6989b
  • 54,422