1
$\begingroup$

Let $\mu\sim N(0,1)$, $Z\sim N(\mu,1)$. Then $Z$ can be viewed as a mixture of Gaussians. It can also be viewed as a Gaussian but there is a prior for the mean.

Let $X\sim\exp(\lambda)$ where the density of $\exp(\lambda)$ is parameterized as $\exp(-x/\lambda)/\lambda$. The independence of $X$ and $Z$ is also assumed. Now I'm interested in the asymptotic behavior of the following probability: $$\mathbb{P}_\lambda(Z^2\leq X)$$ as $\lambda\to0^+$.

Here $Z^2$ can also be viewed as a non-central chi-squared distribution with 1 degree of freedom and a prior distribution of $\chi_1^2$ on the non-central parameter.

Obviously this probability tends 0 as $\lambda\to0^+$. I also did some simulation study and it seems that the order should be $\sqrt{\lambda}$. However, I'm not able to come up with a rigorous proof and have no idea on where to start.

Simulation

$\endgroup$
  • $\begingroup$ Nope. $\lambda$ is another constant and $\mu$ is the Gaussian random variable as defined. Thanks! $\endgroup$ – neverevernever Sep 2 '18 at 19:38
1
$\begingroup$

The probability can be computed exactly as $$ \int_{\mu=-\infty}^\infty \int_{z=-\infty}^\infty \int_{x=z^2}^\infty \frac{e^{-\mu^2/2}}{\sqrt{2\pi}} \frac{e^{-(z-\mu)^2/2}}{\sqrt{2\pi}} \frac{e^{-x/\lambda}}{\lambda} dx\, dz\, d\mu = \sqrt{\frac{\lambda}{4+\lambda}} $$ So this looks like $\sqrt{\lambda}/2$ as $\lambda\rightarrow 0^+$.

$\endgroup$

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.