Looking at cases when early burst can be discriminated from OU, a la Harmon et al 2010 (Harmon et. al. 2010). Fixed models can be told apart. Rather trivially, none of these models can be told apart if initially estimated from Brownian data. Rather encouraging: models estimated from common data can be told apart. 600 taxa tree
Still works on 60 taxa, identifying OU correctly when sim under OU (left), and EB for EB data (right):
But not always.
Parameter estimation on weak EB can be quite tough. 600 taxa tree does fine with a stronger EB signal, though with considerable uncertainty
For instance, the small geospiza data set (13 taxa) the OU fits better, but not substantially:
Should add parameter distributions for the above cases and run the model choice on a couple simulations with weaker EB parameters, as well as a couple of the real data sets.
References
- Harmon L, Losos J, Jonathan Davies T, Gillespie R, Gittleman J, Bryan Jennings W, Kozak K, McPeek M, Moreno-Roark F, Near T, Purvis A, Ricklefs R, Schluter D, Schulte II J, Seehausen O, Sidlauskas B, Torres-Carvajal O, Weir J and Mooers A (2010). “Early Bursts of Body Size And Shape Evolution Are Rare in Comparative Data.” Evolution. https://dx.doi.org/10.1111/j.1558-5646.2010.01025.x.