Let $\xi_p$ be pth quantile of the distribution $F(x)$ with derivative at $\xi_p$, $f(\xi_p) >0$. Then,

$$ |\hat\xi_{p,n} - \xi_p| \leq \frac{2}{f(\xi_p)}\sqrt{\frac{\log n}{n}} $$

almost surely for large enough $n$ and any $p \in (0,1)$, where $\hat\xi_{p,n}$ is the pth sample quantile from an empirical distribution $F_n(x)$ for $F(x)$ with $1/n$ mass for each observation.

I have tried Taylor expansion to get a bound but this approach does not seem to work; it seems like I need to rather work with empirical distribution (using Hoeffding's inequality) rather than sample quantile to derive the term $\log n$, but I am not sure of how I can derive this upper bound.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.