Testing the ABCIMA Code

That’s exactly what’s happening right now.I’m killing time while the code runs. It’s really slow. So far the only bugs have been a couple typos and forgetting that R’s Single Value Decomposition function returns the diagonal elements of the diagonal matrix as a vector not as a matrix. We’ll see what else crops up in the mean time.

The really sad part is that ESS locks up emacs when waiting for code to run, so that means I can’t use my favorite text editor to work on other project related code, like some of the degeneracy checks and simulation scripts previously discussed.

Speaking of the diagonal part of the single value decomposition, there’s an interesting interpretation which can be drawn from, strangly enough, dynamical systems theory. In that theory, when we diagonalize a system the eigenvalues represent which direction a point gets pushed along a given eigendirection. In a statistical context, that would correspond to how strong the dispersion is along a particular axis of the bivariate normal distribution. Taking samples on a particular line, then are going to be more spread where the eigenvalues are larger but not necessarially down some particular arbitrary axis. It’s kind of cool.