Adjustable distance prior

[karllark]

Adjustable distance prior

[drvdputt]

In case someone starts working on this, I want to bring up the following:

The beast has some code in grid_weights.py to deal with irregular grid spacings. E.g. for Z, the prior in bin-space is set equal to the the bin widths (based on the middles between the grid points in linear space). When summing over the grid, this corresponds to a flat prior in Z-space (sum becomes an integral)

When I implemented the distance parameter, I did not apply these weights to the distance (d), and so the weight is 1 for each bin, so the prior is flat in bin-space. So in d-space, we currently always have a 1/bin_width prior.

For a log-spaced distance grid this makes sense. If you do the math for this specific case, you can find that the bin width (however you calculate it) is a constant times the center value of the bin. So the prior is 1/d in d-space, which is flat in log(d)-space.

In any case, it would be better to just apply the same correction as for Z, and have a 1/d prior explicitly as an option.