I have the following marginal posterior of a vector $\phi$ ($p$ by $1$):

$$p(\phi | Y) \propto \left[1+\frac{1}{h}\left(\phi - \tilde{\phi} \right)' \Gamma \left(\phi - \tilde{\phi} \right) \right]^{-\frac{n+p}{2}}$$

where $Y$ denotes the data, $h$ is a constant, $\tilde{\phi}$ ($p$ by $1$) is a vector of constants, $\Gamma$ is a $p$ by $p$ matrix of constants, and $n$ is a constant which is different from $h$.

I'm told that the marginal posterior of $\phi$ is a multivariate $t$-distribution, however, checking wikipedia (http://en.wikipedia.org/wiki/Multivariate_t-distribution), although it appears to be almost the same as the multivariate $t$-distribution, there is a small difference. Namely, the $\frac{1}{h}$ should be a $\frac{1}{n}$, but here $h$ and $n$ are different.

Could someone clarify this?


1 Answer 1


So ... if you take a factor of $n/h$ into $\Gamma$ ($\Gamma^*=\frac{n}{h}\Gamma$, say), what does it look like then?

  • $\begingroup$ Ah, can't believe I didn't see that haha, thanks @Glen_b! $\endgroup$
    – TrueTears
    Jan 12, 2015 at 12:48

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.