I am reading a chapter on concentration inequalities, and I am struggling to make connections on sub-Gaussian random variables. A random variable is sub-Gaussian if there exists $C > 0$ such that $P(|X| \geq t) \leq 2\exp(-t^2/C^2)$. The text gives numerous equivalent conditions to this one.

I have an exercise: If $X \sim Poi(\lambda)$, show that $X$ is not sub-Gaussian.

My setup is $P(|X| \geq t) = P(X \geq t) = \sum_{k=t}^{\infty} \frac{e^{-\lambda}\lambda^{k}}{k!}$ since the Poisson random variable is supported on nonnegative integers. My first thought was to use Stirling's formula on the $k!$, but I wouldn't know how to sum the resulting series. I appreciate any help on this question!



I don't think you need to do any summing. You have $$ P(X\ge t) > \frac{e^{-\lambda}\lambda^t}{t!}.$$ You can apply Stirling to this to get an asymptotic lower bound and show it decays slower than any Gaussian.


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.