Another Look At Confusing Moves

And so I went back to the code . . .

Long story short, plenty of points spread all over the feasible region are transitioned through. As the previous post notes, many of the expected ones are ignored by the Guassian Transition method opposed to the uniform (which transitions quite nicely). After looking at the code (for the 10th time), the only difference is the ratio of the proposal densities.

So what does this mean? It should mean that when the mean of the proposal is far from the true value, this ratio will help drive the chain towards the true value and vice versa.

Time to figure out why my code is doing the opposite.