Lab Notebook: August 1

Still working on paper writing, unfortunately, but I met with Ryan, Dave, and David today re: rate partitioning. There are basically two upsides to partitioning data sets by rate:

1. It appears to work well for certain datasets (the Gobies). We hypothesize that this is because of several splitting events occurring in deep time, but we still haven't simulated data (or found a similar dataset) that would confirm this.
2. Tree inference works faster. Partitioning the data in this way seems to make it clear to the tree building programs that certain partitions are better for certain splits. This appears to allow the tree inference algorithm to more easily discard sites that are not informative for the tree. It has a similar effect as placing a prior: these are the data we know are good for resolving this tree, so use it.

Here's what we want to do now:

1. Get McCormack data
2. Rate partition
3. Timed tree inference