Let $p, q \in \mathbb{N}$ and $Z$ a gaussian process over $D \subset \mathbb{R}^{p+q}$ : $Z(\cdot) \sim \mathcal{GP}(\mu(\cdot), k(\cdot,\cdot))$.

For $(x_1, \ldots, x_p, y_1, \ldots, y_q)^T \in D$, we define an associated process $W$ as

\begin{align} & W(x_1, \ldots, x_p, y_1, \ldots, y_q) \\[10pt] = {} & \frac{\exp[Z(x_1, \ldots , x_p, y_1, \ldots, y_q)]}{\int_{u_1}\cdots\int_{u_q} \exp[Z(x_1, \ldots, x_p, u_1, \ldots, u_q)] \,du_1 \cdots \,du_q}. \end{align}

How to find the distribution of $W$ ?

I considered the following lead : if $Z$ has continuous paths, then we can see the integral as the limit of a Riemann sum. However, I found that there was no closed-form formula for the distribution of a sum of dependant log-normal variables...

I would be grateful if some of you could give me some hints, other leads I can follow or for any related answer :)

Thanks in advance !

  • $\begingroup$ Maybe this would be better at Cross Validated, the stats version of mathoverflow $\endgroup$ – David White Feb 6 at 15:32
  • $\begingroup$ It seems quite reasonable for MO given the probabilists we have here. Perhaps the downvoter finds the question easy? $\endgroup$ – Yemon Choi Feb 8 at 0:20

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.