Peter Ralph
Peter Ralph
Great, @nspope - this is perfect! Well-explained, and I think also a very good approach. Here's some notes: - The numbers `x` and `k` are the "raw" input here, and...
Nice idea, @nspope - but, since nodes extend over many trees and the number of samples below them can change, the appropriate entry in this table wouldn't be the node,...
> I suppose what I'm suggesting is an interface to get summary statistics from any time-to-event distribution that arises from counting a particular topology in a tree sequence for each...
> I still think it's possible to avoid multiple rows per node (because the total weight for a given node is the sum of weights for that node across marginal...
I think if we get one or two more good examples besides "coalescence rate" then we'll see how to generalize. > I am suggesting we keep track of the left-most/right-most...
I'll be curious to see the details - there's a lot of possible ways to go with the details.
My two cents (mostly, for not complicating things): > 1. Should max_time include the max time of mutations too? Sure, why not. > 2. do we count unreferenced non-sample nodes...
Hm - notes: - `update_node_summary` and `update_running_sum` can be combined into one (I think), eliminating some extra copying around of results - it's a bit hard to know how well...
ping @jeromekelleher
Yeah, you're right. I guess the main question is whether to make `samples` optional with default `None` or make people pass in `None` explicitly (the first, I imagine?).