Let $\mathbf{X}$ be a vector-valued random variable with finite second moment and density $\rho$. Assume that $\rho$ is bounded and continuous. As $\mathbf{X}$ has finite second moment, I hope to find a bound of its density $\rho(\mathbf{x})$ for large $\lvert\mathbf{x}\rvert$ by a density of Gaussian type. My question is, more specifically, if it is possible to find a covariance matrix $\Gamma>0$ and a constant $M$ such that for large $\lvert\mathbf{x}\rvert$, it holds that $$\rho(\mathbf{x}) \leq M\pi_{\Gamma}(\mathbf{x}),$$ where $\pi_{\Gamma}$ denotes a normal density with a suitable mean and covariance matrix $\Gamma$. Counterexamples or oppositions are, of course, also welcome.

Thank you!


Why do you think this should hold ? In the one-dimensional case, consider $\rho$ the density of an exponential distribution $\mathcal E(1)$. Since for any $a$, $\exp(-(x-a)^2/2) = o_{+\infty}(\exp(-x))$ it's not possible to find the $M$ you're looking for.

| cite | improve this answer | |
  • 1
    $\begingroup$ Well, sometimes you are too much biased since you hope for something :P. All the mentioned cases make sense and are, kind of, obvious. Thank you! $\endgroup$ – MTP May 22 '19 at 6:24

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.