Brown and Yang (2011; http://www.biomedcentral.com/1471-2148/11/271) show that data generated with S < 0.1 can be effectively modelled using a strict molecular clock. So we should have a prior on S that gives 50% weight below 0.1 and 50% weight above (i.e. a gamma prior with median of 0.1 and a reasonable spread, up to about 1.0).
Proposed prior for S:
gamma(shape=0.5396, scale=0.3819)
has median of 0.1 and 97.5% probability below S=1.
Then a heuristic test of clock-likeness is just whether the posterior estimate of S < 0.1.
The BEAST2 book discusses Brown and Yang (2011) and the 0.1 cutoff in chapters 9 and 10, so we should match the software with that discussion.
Brown and Yang (2011; http://www.biomedcentral.com/1471-2148/11/271) show that data generated with S < 0.1 can be effectively modelled using a strict molecular clock. So we should have a prior on S that gives 50% weight below 0.1 and 50% weight above (i.e. a gamma prior with median of 0.1 and a reasonable spread, up to about 1.0).
Proposed prior for S:
gamma(shape=0.5396, scale=0.3819)
has median of 0.1 and 97.5% probability below S=1.
Then a heuristic test of clock-likeness is just whether the posterior estimate of S < 0.1.
The BEAST2 book discusses Brown and Yang (2011) and the 0.1 cutoff in chapters 9 and 10, so we should match the software with that discussion.