MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am studying the (asymptotic) behavior of the p.d.f of the random variable $Y = X + Z$, where $X$ is an r.v. with any distribution function $F(x)$ such that $\int_{-\infty}^{\infty} x^2 \mathrm{d}F(x) \leq P $, and $Z$ is independent of $X$, with a Gaussian density

$p_Z(z) = \frac{1}{\sqrt{2\pi\sigma^2}}e^{-\frac{z^2}{2\sigma^2}}$.

So the density of $Y$ is the convolution $p_Y(y) = \int_{-\infty}^{\infty}\ p_Z(y-x)\ \mathrm{d}F(x)$.

Applying Jensen's inequality, we know the tail of $p_Y$ decays no faster than a (scaled) Gaussian function with some mean and variance $\sigma^2$. Also by bounding $y^2p_Z(y-x)$, we see that $p_Y$ decays no slower than $cy^{-2}$. Of course the tail of $p_Y$ depends very much on $F(x)$. I am wondering in which area people study such relations, and what literature I should look for. I know in image processing, people use "Gaussian kernel smoothing/filtering" methods, but they are more interested in the algorithms etc.



P.S. (in response to Yuri) Currently I am able to analyze the tail of $p_Y$ in some specific cases. For instance, if both ends of $F(x)$ decay faster than any Gaussian, then the tail of $p_Y$ is roughly the same Gaussian as $p_Z$. On the other hand, if $X$ has a density that decays slower than any Gaussian and is not changing too fast (e.g. reciprocals of polynomials, or exponential), then $p_Y$ has the same tail as the density of $X$. However, these regular examples are not general enough to represent all distributions $F(x)$. For example, if the tail of $X$ is some times Gaussian, sometimes polynomial/exponential, then what can we say about the tail of $p_Y$?

share|cite|improve this question
Fourier transformation gives $\hat{p_Y}(\xi) = \hat{p_Z}(\xi) \hat{dF}(\xi)$, where $\hat{p_Z}$ is a Gaussian, so in the frequency domain you have exact control of the influence of $\hat{dF}$. Perhaps you can then use Fourier analysis to draw conclusions for the asymptotics of $p_Y$. – Patrik Wahlberg Nov 20 '11 at 11:24

If $X$ has polynomial tails, then so does $Y$.

share|cite|improve this answer
Thanks. Here the length of response is too limited, please see the P.S. part of my question. – Wei Mao Nov 20 '11 at 18:22

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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