Plot of edge areas, possibly with corresponding mutation density

[hyanwong]

It is useful to look at the distribution of edge "areas" (i.e. (parent_time-child_time) * span: note this may not be quite so meaningful in an undated tree sequence, but might still reveal some issues).

In particular, if we have mutations on the tree sequence, we expect the number of mutations to be proportional to edge area. If there are edges in the tree sequence that cover a large area with hardly any mutations, this is likely to indicate QC issues. Such regions, by the way, also cause problems for tsdate (causing it to try to estimate some ancestral nodes in the GEL data as coming into existence only a few seconds ago). We seem to be seeing this with very long recent edges: perhaps due to a lack of singletons?

As a guide, perhaps it would be useful to have a plot where it is obvious if (say) half the singletons were removed.