1

Let $X \sim \mathcal{N}(M, S)$, where $M,S$ are themselves gaussian random variable with mean $\mu_{M,S}$ and variance $\sigma_{M,S}$. Does this distribution have a particular name/form? Can we compute its CDF?

jojo87
  • 45

1 Answers1

0

If S is Gaussian then it could be negative which does not work. This problem is common in Bayesian statistics. Possibly specifying the precession $=1/S^2$ will work better. It will be easy to integrate out the M, this will just give a new Gaussian where the mean depends on $\mu_M$, $S$ and $\sigma_M$. Integrating out the S will be messy and possibly not possible.

user121049
  • 1,599