Let $W(t)$ be an $n$-dimensional Wiener process. Denote by $\chi_n^2$ a chi-squared random variable with $n$ degrees of freedom. I have recently found the following inequality given without proof: $$ {\mathbf P} \left\{ \max_{t\in[0,2]} \|W(t)\|> x\right\}\leqslant 2 {\mathbf P} \bigl\{\chi_n^2>x^2/2\bigr\}. $$ Why is this bound true? What other deviation inequalities for the Bessel processes are known?

  • $\begingroup$ This should be very closely related to the maximum process of Brownian motions, see e.g. Proposition 13.13 in Kallenberg's Foundations. $\endgroup$ Jan 8, 2014 at 7:38
  • $\begingroup$ Yes, it is related to the reflection principle. In fact, we should better work with the probability for the process $W(t)$ to exit a sphere of radius $x$. Then we can modify the proof for one-dimensional Brownian motion so that the bound holds. Thank you. $\endgroup$
    – Raindog
    Jan 8, 2014 at 10:44

1 Answer 1


A general bound that's sometimes useful is given by Fernique's theorem. It's a general fact about Gaussian measures on Banach spaces, but in this case it gives the following: there are constants $C, \epsilon > 0$ (depending on the dimension $n$) such that $${\mathbf P} \left\{ \max_{t\in[0,2]} \|W(t)\|> x\right\}\leqslant C e^{-\epsilon x^2}.$$

You can find a proof in these lecture notes of mine (see Section 4.2).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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