A zero mean sub-Gaussian random variable Z satisfies $Eexp(tZ)≤exp(t^2k^2/2)$ for some constant k>0.

Let's say we have random variables: $X_1,X_2,...,X_n$ with $EX_i=\mu$ and $Var(X_i)=\sigma^2$. We know that $\hat{\sigma}$ is consistent. I am interested in obtaining tail bounds for the inverse of the estimator of the standard error: $1/\hat{\sigma}$, where $\hat{\sigma}=\sqrt{\frac{1}{n}\sum_{i=1}^n (X_i-\bar{X})^2}$ and $\bar{X}=\frac{1}{n}\sum_{i=1}^nX_i$.

I didn't find anything in the literature.

If the $X_i$ are standard Gaussian random variables, then $\hat{\sigma}^2$ is distributed according to a chi-squared distribution. Can we prove that if $X_i$ are sub-Gaussian, $\hat{\sigma}$ is sub-Gaussian or sub-exponential? And what about its inverse?

Or more generally, can we prove that under some assumptions on the $X_i$, it exists a $n$ such that $1/\hat{\sigma}$ is sub-Gaussian?

Thanks you in advance


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.