The calculation of tau is not correct

[ferrine]

https://github.com/pymc-devs/pymc-experimental/blob/dd04eff3e1e1b3de88e8f20ad2ad34ce01402bb4/pymc_experimental/distributions/multivariate/r2d2m2cp.py#L68

In case $R^2$ is close to 1, tau goes to infinity, which makes zero sense for me. I have feeling that tau should be $\tau=R^2$

[ferrine]

My parameterization was slightly different, and I think that in my case there is no tau since I use output variance information to create all the parameters