I simply want to fit multivariate regression with multiple outputs to extend the linear regression tutorial.

Let’s say Y is a vector. Could you show me how to adjust shape in model.

Are you assuming that the outputs have correlated errors? Or do you want to have several multiple regressions stacked in vector format?

The latter. One way is to have multiple regressor each target. But I want to know how to extend GLM to multiple targets output regression.

I don’t think I understand. Can’t you just stack your outputs as a single vector, then code one big regression with different covariates (including intercepts) for each effect per regression?