lkirk

After talking with @apragsdale, I think that the LD interface will differ from the other stats interfaces. My understanding is that tskit typically groups stats by name and allows parameters...

@jeromekelleher I didn't quite see the subtlety between the last two you proposed until just now, specifically, I didn't see the `positions`/`sites` mode. I had assumed that users would provide...

I've had a couple of competing thoughts... ### Explicitly listing the sites Advantages: we can leave one out easily (i.e. `right_sites=[4, 6, 7]`) Disadvantages: the list is longer ``` ts.r2_matrix(left_sites=[0,...

@jeromekelleher I think we that it depends on the way we want to define these. If we use ranges like you've suggested, then we can have a single `sites` arg...

@jeromekelleher I see, I thought you were talking about ranges, not explicitly addressing each site. I agree that `left` and `right` are not useful terms to describe arrays. I think...

Yes, I agree. I think removing windows entirely and layering them back in as needed is probably the way to go. I can add them when we get to LD...

@jeromekelleher I think that the reorganization sounds like a great idea. I agree with what you've proposed.

@jeromekelleher thanks for pointing me to the proper usage. I have gone ahead and modified the code to properly consume the `seek_forward` result. > **However** there is a subtle issue...

Hi @jeromekelleher sorry for my delayed response. I did a little cleanup, it's rebased and squashed now. Ready for merging.

When we last met, we had discussed: - explicitly testing output dimensions. I suppose I misinterpreted the goal. As mentioned, I'm testing the number of dimensions and not the actual...