Picking exemplars

[rvosa]

When picking the shallowest exemplars for the backbone, exemplar clades where shallow tips have near zero distance to their subtree root result in a backbone where nearly all change is allocated to the root branch of the of the exemplar clade. When grafting the subtrees onto the basal split, the subtrees become "too tall". Here's the grafted example:

Here's the same topology but with the branch lengths re-estimated by raxml:

A follow up experiment should be to pick the tallest exemplars instead. It's possible that this will make them "too shallow" instead, especially when there is a lot of homoplasy within the clade.

[rvosa]

Note also that the clade that sticks out very far in the re-estimated tree is rooted incorrectly, which needs to be fixed as per issue #76

[rvosa]

The way to go about this is as follows:

• create an exemplar variable in the config.yml, which can be tallest, shortest, median
• also create a scaling variable, which can be exemplars or outgroup, i.e. scale by the height ratio of the crown in the backbone over the ingroup crown, or the ratio of the distance to the outgroup in the backbone versus that in the subtree
• implement the various scenarios with case/switch statements in the exemplar picking script and the grafting script
• compare all possibilities with re-estimated branch lengths: which minimizes the R^2?

[rvosa]

Picking this up in #79