Got it. For the diagonal covariance case that explains why it's just the square root of the matrix, which are of course the std deviations.
You don't necessarily have to implement a constructor that takes in the covariance matrix. The ones that take the Cholesky decomposition could just be made private as they are not otherwise used yet. Heck you don't even need a matrix multiplication if not dealing with the off-diagonal elements. On Wed, Nov 14, 2012 at 5:58 PM, Ted Dunning <[email protected]> wrote: > On Wed, Nov 14, 2012 at 9:48 AM, Sean Owen <[email protected]> wrote: > > > I'm talking about the case here where covariances are 0. The marginals in > > each dimension are independent and are normally distributed. Right? > > > > Yes. With no covariance, all of the axes are independent. > > > > What is that matrix connecting the univariate normals to multivariate > BTW? > > > > > http://en.wikipedia.org/wiki/Multivariate_normal_distribution#Drawing_values_from_the_distribution > > > > I don't know what it is theoretically. It's not the covariance matrix, > > which might be a reader's first guess on looking at the constructor. > > > > Should I fix the constructor to accept covariance? The SVD or Cholesky > decomposition required is pretty quick. >
