I have a random variable with an unknown distribution and I want to find its cumulative distribution function. I sample the distribution $N$ times, with $$X_1, \dots, X_N$$ being iid random variables. I then plot the empirical cumulative distribution function.

Using this article https://en.wikipedia.org/wiki/Dvoretzky%E2%80%93Kiefer%E2%80%93Wolfowitz_inequality

Am I correct in assuming that if I let $N$ be such that $$N \geq \frac{1}{2\epsilon^2}\ln(\frac{2}{\alpha})$$ Then my empirical distribution function is within $\epsilon$ of the actual cumulative distribution function at all points with a $1-\alpha$ confidence?


Yes, if $N \geq \frac{1}{2\epsilon^2}\ln(\frac{2}{\alpha})$, then $$ \mathbb{P}\left(\sup_{x \in \mathbb{R}}|F_N(x) - F(x)| > \epsilon \right) \le 2 e^{-2N\epsilon^2} \le \alpha, $$ which is what you want.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.