Let $\mathbf{X}$ be an $m\times m$ random matrix full rank matrix, having the density function $f_{\mathbf{X}}(X)$. Also, let $\mathbf{W}$ be a deterministic $k\times m$ matrix of rank $k$ and $k<m$. What is the distribution of $\mathbf{Y}=\mathbf{W}\mathbf{X}\mathbf{W}^T$? Basically my question is how to calculate the Jacoubian $J(X→Y)$.
Asked
Viewed
209 times
1

$\begingroup$ unlike in your previous question,  mathoverflow.net/questions/169041/…  you now have a linear transformation, so the Jacobian is constant and only affects the normalization of your distribution; so you don't really need it, but it's $$\left\left\frac{\partial WXW^{\rm T}}{\partial X}\right\right={\rm det}\,(WW^{\rm T})$$ $\endgroup$ – Carlo Beenakker Jun 12 '14 at 20:13

$\begingroup$ Thanks Carlo. I still can not write the distribution of $\mathbf{Y}$ since the transformation from $\mathbf{X}$ to $\mathbf{Y}$ is not invertible. What is $f_\mathbf{Y}(\mathbf{Y})$? $\endgroup$ – Peter Jun 12 '14 at 23:07

$\begingroup$ you'll have to integrate over the degrees of freedom of $X$ that do not appear in $Y$; without knowing $f_X(X)$ there's no way to tell you what the outcome of the integration will give. $\endgroup$ – Carlo Beenakker Jun 13 '14 at 0:48