You can use something like corHMM (https://doi.org/10.1093/sysbio/syt034) and perhaps its rayDISC function to find parts of the tree where the transition rates between states are different (for your question, where rates might be higher). Basically, it allows a hidden "trait" (which isn't necessarily a single trait, just some thing or things that change on the tree) to affect the transition rates. A good example is https://academic.oup.com/view-large/figure/115233499/syt034f3.jpeg -- you can see where transition rates are higher. The nice thing (though note my conflict of interest, as a coauthor) is that it allows the rates to actually change over the tree; different approaches that use a single rate matrix over the whole tree and then reconstruct changes (using things such as stochastic character mapping) are essentially using a model to infer things in order to find deviations from a model. One downside of the corHMM model is that the hidden changes evolve under a continuous time Markov model -- fine if you think whatever factors affect character rate (biogeographic changes, presence of certain other species, new morphological traits, etc.) evolve in this way (which is true for nearly all of them) but not if a single factor appears suddenly (i.e., radical change in environment for all species once an asteroid his the planet). CorHMM might have a timeslice function to allow this, but I'm not sure.
If you have a lot of characters, a different approach could be to estimate branch lengths under something like a Lewis MKV model, which basically gives you amount of change averaged across all characters, and then compare the ratio of those branch lengths to the chronogram branch lengths for those same edges. Those branches with higher than the average ratio have more character change per unit of time than other branches. There's probably a paper or papers that do this, but I don't recall any at the moment, my apologies. If the question is about two rate regimes (i.e., before or after a massive event) or gradual change of rates, you could do tree transformations: use Geiger's tree transformation function with discrete characters to find the transformations that best fit the data -- maybe more change earlier than later, a discrete shift in rate 30 MY ago, etc. Some standard caveats on these and other potential solutions: ancestral state reconstruction is very hard to do well: N tips have N-1 ancestral states to estimate, plus rate parameters. That's a lot, which is why methods that fit overall rates could be better than reconstructing many changes. But it's possible to over-interpret these, too: you can in theory from corHMM get an estimate of the support for every hidden state and rates for every node, but I wouldn't read a lot into every single node. Also, there's still the challenge of Maddison & FitzJohn (2015): https://doi.org/10.1093/sysbio/syu070. It still scares me, personally, and the issues raised affect our interpretation of many methods that look at rates, such as the ones here. Best, Brian _______________________________________________________________________ Brian O'Meara Professor, Dept. of Ecology & Evolutionary Biology, UT Knoxville Associate Head, Dept. of Ecology & Evolutionary Biology, UT Knoxville He/Him/His On Thu, Apr 9, 2020 at 2:55 PM Elizabeth Miller <emil...@uw.edu> wrote: > I am interested in testing whether ancestral state changes (e.g. habitat > transitions) are clustered in time and/or in specific lineages rather than > randomly distributed on phylogeny. Does a method already exist for testing > this? I am aware of BayesTraits for testing correlation among states, but > not a method for clustered distribution of state changes with time. What > immediately comes to mind is simulating an expected temporal distribution > of state changes if changes were random and comparing to observed temporal > distribution, but I just want to check if something more sophisticated > exists. > > Thank you! > > -- > Elizabeth Miller, Ph.D. > NSF Postdoctoral Research Fellow > School of Aquatic and Fishery Sciences > University of Washington > > [[alternative HTML version deleted]] > > _______________________________________________ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at > http://www.mail-archive.com/r-sig-phylo@r-project.org/ > [[alternative HTML version deleted]] _______________________________________________ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
