1
$\begingroup$

I try to solve a gaussian convolution which turns out to be harder than expected. Here is the isolated difficulty: \begin{align} \int_c^{+\infty}\phi(x-y)\big(1-\Phi(y)\big)dy \end{align} where $\phi$ and $\Phi$ are the pdf and cdf of the standard gaussian function.

My idea was to write explicitly the Heaviside and to use the integral property of convolution. But I don't know whether it is correct in this case:

\begin{align} &\int_c^{+\infty}\phi(x-y)\big(1-\Phi(y)\big)dy \\& =\int_{-\infty}^{+\infty}\phi(x-y)H(y-c)\int_y^{+\infty}\phi(z)dzdy \\ &= \int_{-\infty}^{+\infty}\phi(x-y)H(x-y+c-x)\int_{-\infty}^{-y}\phi(z)dzdy \\ &= \int_{-\infty}^{+\infty}\phi(\bar{x}-\bar{y})\bar{H}_{x-c}(\bar{x}-\bar{y})\int_{-\infty}^{\bar{y}}\phi(z)dzd\bar{y} \\ &= \int_{-\infty}^{\bar{x}}\int_{-\infty}^{+\infty}\phi(z-\bar{y})\bar{H}_{x-c}(z-\bar{y})\phi(\bar{y})d\bar{y} dz=\int_{-\infty}^{\bar{x}}(\phi\bar{H}_{x-c})*\phi(z) dz \end{align} which I can compute. But is the passage between the two last lines correct? Is there another way? Here $\bar{x}=-x$ and for the Heaviside $\bar{H}(t)=H(-t)=1-H(t)$.

$\endgroup$
1
$\begingroup$

Some informations about this puzzle, if anybody ever meet this situation:

This integral is not documented in famous tables of integrals, which is not a good sign. Most probably, this integral can not be expressed in a simpler closed form.

This integral arised in a probability problem. I wanted to find the distribution of a random variable defined as a sum of two other random variables with known distributions. I have ended up in this dead end convolving them without noticing that they where in fact not independent variables.

$\endgroup$

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.