README or elsewhere discussion of alternatives

[seabbs]

Hi,

Me again! Another little feature request.

I would really like to see a small section on alternatives and how this differs from them. That could be scoring approaches commonly used in ID or it could be a comparison to other packages for comparing trees.

[CyGei]

@seabbs Thanks for the suggestion! Our preprint is available here. In the Discussion we briefly mention the GTST Python package, which relies on abstract graph kernels. These are not directly suited to our setting, although GTST appears to allow custom distances. The potential advantage of PERMANOVA seems to lie in its ability to quantify how much variation in graph collections is due to other variables e.g. inference method, data type, and prior assumptions, which I do not think GTST supports (I may be wrong!)

I see two next steps that may be useful and would be happy to discuss further if of interest?

1. Perform a simulation study that compares the performance of GTST with our PERMANOVA approach.
2. Evaluate the ability of PERMANOVA to explain variance in a multifactorial setting.

Could you please clarify re: using scoring approaches?

[seabbs]

Yes I read that thanks! I think either of those might be a good idea but what I am really getting at here is that it is useful in package docs (the entry point for most people) to clearly tell people what the alternatives are. Here the main alternatives are other tree distance measures and more traditional epi scoring methods for non-tree data I would guess.