Back in December, I put a paper on the arxiv about Fisher matrices. You can see my blog post about it here. I've since put together a short Jupyter notebook detailing how we actually went about calculating each Fisher matrix, which you can see here.
I was using quite an old version of COMPAS for the results in the paper, where the output format wasn't written very well, thus all of the horrible bookkeeping code. I've since rewritten that part of COMPAS to enforce some better standards.