Let $\Sigma$ be a $p\times p$ positive definite matrix and $\nu>p-1$. I would like to simulate a realization of the Inverse-Wishart distribution $X\sim\mathcal W^{-1}(\Sigma, \nu)$ with the constraint that certain off-diagonal terms of $X$ should be 0. For example, the constraint could be that $X_{1,p}=X_{p,1}=0$.

In practice, this will be useful in the following Bayesian setting: $X$ will be the covariance matrix of a multivariate normal distribution for which I know that certain components are independent. This will be used in an MCMC, in which the parameters $\Sigma$ and $\nu$ change at every iteration, so I am looking for a reasonably efficient way of getting a single realization. The dimension $p$ will typically be between 3 and 15.

I would appreciate any pointers on how to handle this problem.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.