Matthew Miles Osmond

This looks like the right thing to me. One semantic note though, should the new method you made be called ```genetic_relatedness_noncentred_summary_func``` rather than ```genetic_relatedness_centred_summary_func```? I was trying to get this...

Thanks @petrelharp -- it would have taken me a long time to figure out that format specifier. Yes, I'm interested in doing this tree by tree. And yeah, a time-window...

> how will you do this without a time window? I'd calculate the non-centred shared times with ```genetic_relatedness```, write a little script to convert the output to the shared times...

Thanks @petrelharp! > So if a branch is ancestral to neither sample, then it counts also. Ah, now I get it, thanks. > Somehow this does the right thing for...

> Hm, I'm not sure this works out - each branch only counts in one direction, since for the case of single nodes, if x1 * x2 is nonzero then...

> And, oh! I see about the epochs. It only works if you're working with only single samples on a single tree and all samples are at the same time,...

Yeah, a hand would help. I took a stab at fixing the failing tests but can't quite piece it all together. It looks like the problem is passing the extra...

> Shall we change it to centre? I was thinking the same and just committed the changes to ```_tskitmodule.c``` and ```trees.py``` (the ```trees.c/h``` files still refer to a ```noncentred``` option/summary...

Hey @petrelharp, sorry, having a hard time finding the time right now (and it takes me a while to navigate the code). If you can find the time go for...

See #23 -- I think you need to use python2 (it worked for me)