Equation 3.26

[hckiang]

I think the the expansion of the Kronecker product is "reversed": the explicit matrix expression you wrote should be C kronecker R rather than R kronecker C.

If we take X as a one-observation-per-role matrix then the covariance (of vec(X)) should be R kronecker C.