0
$\begingroup$

I need to take independent draws from Normal Inverse Wishart distributions with different parameters. In practise I want to draw:

$x_i|\Sigma_i \sim \mathcal{N}(\mu_i, \Sigma_i \otimes \Psi_i )$ $\Sigma_i \sim IW(S_i, v)$,

for $i = 1 , \ldots, T$

where $\Sigma_i$ is $N \times N$ and $x_i$ is $k \times 1$. The parameter of the degrees of freedom of the Inverse Wishart is the same for all the distributions.

Right now, I am doing the draws from the Normal Inverse Wishart distributions one by one, and looping over T. I am using this Matlab code:

    for i = 1:T
      z = chol(inv(S{i})'*randn(N,v);   
      sig_sim = inv(z*z');
      vc = kron(sig_sim,Psi{i}); 
      [cf,pp] =chol(vc);
      x_sim  = mu{i} + cf'*randn(N*k,1); 
    end

Is there a way to avoid the loop over T and draw all the T Normal Inverse Wishart jointly (in a sigle step), since they are independent?

I am not asking how to take the draw from the marginal distribution of $x$. The question is not related in any way to that. I want to understand how to take a draw for $\Sigma_{1:T}$ and $x_{1:T}$ without looping over i.

Improve this question
$\endgroup$
2
  • $\begingroup$ you should read carefully the questions before closing them! $\endgroup$
    – Giorgetto
    Nov 1, 2022 at 18:15
  • $\begingroup$ If you are still interested, you can edit the post to explain why the proposed dup do not answer your Q. It will then enter a queue for reopening! $\endgroup$
    – kjetil b halvorsen ♦
    Sep 10 at 20:37

0

Reset to default