Milton, Brian, etc, Just wanted to add some background on Alroy's anc-desc change analyses, and the recognition of these relationships in the fossil record. This may be tangential; it isn't clear to me what sort of dataset that Milton has. So, Brian said...
> In Paleo, you can (well, arguably) see the ancestor and descendent pair, so > you can assess the amount of change and the ancestral state. Yes, in some extremely well-sampled fossil records, it is argued you can qualitatively recognize ancestor-descendant relationships. Diagrams depicting ancestor-descendant relationships are usually called stratophenetic diagrams, even though that term may have been originally defined for something slightly different. This is mostly done in microfossil groups, like plankton; see Aze et al. 2011 for a good example of this in planktonic foraminifera. The quantitative ideal would be that ancestor inference would be part of how we infer phylogenies. Fisher's (1992) stratocladistic method included the potential for ancestral taxa in analyses that included stratigraphic debt; very recently, Gavryushkina et al. (2014) presented a similar model-based method for BEAST2 that included ancestor-inference for tip-dated phylogeny inference with fossil taxa. I have not seen anyone use either stratocladistics or Bayesian tip-dating methods to get ancestor-descendant contrasts yet. More commonly though, ancestor-descendant relationships are recognized in less direct ways, after the phylogeny is inferred. Norell (1992) and Smith (1994) suggest that we can treat earlier-appearing, plesiomorphic taxa as ancestors to their later-appearing, apomorphic sisters on a cladogram, but this ignores the potential for character reversal (Wagner, 1996). Wagner (1995a) applied a confidence interval approach based on sampling in the fossil record to identify ancestor-descendant relationships among sister taxa on a cladogram and then used these ancestor-descendant pairs to test for changes in the rate of trait evolution between the Cambrian and post-Cambrian (Wagner, 1995b). My cal3 method does something similar, but stochastically suggests a large number of possible ancestor-descendant relaitonships, as there is considerable uncertainty in which taxa are ancestral, given the uncertainties that arise from the incompleteness of the fossil record. Some workers have recently used a combination of ancestral trait reconstruction and phylogenies of fossil taxa to infer ancestral states and then calculate the change between them, bypassing the need to infer which observed taxa might be ancestral taxa. (And, this is indeed suggested by Alroy, 2000.) However, there is considerable uncertainty in ancestral trait reconstruction even with non-ultrametric trees, so I strongly recommend against this procedure except for very specific questions, in very specific cases. Alroy (1998) only had a dataset of Cenozoic mammals from North America containing taxonomic hierarchy information, and no phylogenetic information beyond whatever is conveyed by taxonomy. As I understand it, Alroy infered ancestor-descendant relationships by randomly samples pairs of congeneric species from the mammal fossil record, with the earlier appearing taxon always as an ancestor and the later appearing species as the descendant, drawing pairs until all genera had less than two species left in them. Alroy then bootstraps, repeating this procedure many times over and over. I think Alroy only sampled taxa as ancestors once, and thus avoided the non-independence issues. Finarelli and Flynn (2007; 2009) do something similar as Alroy (1998), but also randomly sample possible ancestor-descendant relationships between higher taxa. For more on interpreting the ancestor-descendant change analyses that Alroy presents in the 2000 paper, you should also see Wang (2005, also in Paleobiology), who touches on the variance issue. -Dave PS: I clearly spend a lot of time thinking about ancestor-descendant relationships... On Fri, Apr 24, 2015 at 1:30 PM, Brian O'Meara <[email protected]> wrote: > In Paleo, you can (well, arguably) see the ancestor and descendent pair, so > you can assess the amount of change and the ancestral state. Using just > info at the tips, you are inferring from N taxa N-1 ancestral states and > 2N-2 rates of change. That seems a lot to ask from data, especially since > the rates and states affect each other and so aren't independent. You're > also using a model that assumes an even rate of change over the tree, so an > ancestor will tend to be between its descendants (exactly between, if you > have a two taxon tree with coeval taxa). You could try more exotic models > that allow multiple Brownian rates or even multiple Ornstein-Uhlenbeck > rates, but there's a limit there, as well. There are Bayesian approaches > that can paint a smear (estimated from post-burnin samples) of rates over > multiple branches, but I'd be careful interpreting them as robust estimates > of rate that are independent of the states. > > If you have separate traits, you could ask does trait X lead to higher > rates (or OU mean, or OU alpha) of evolution in trait Y. Basically this > involves painting a tree with discrete trait X (generalist vs specialist > predator, say) and estimating different rates for the other trait (i.e., > mouth volume) on the branches painted in different states (and seeing if > these rates are biologically and statistically significantly different). > > Best, > Brian > > _______________________________________ > Brian O'Meara > Assistant Professor > Dept. of Ecology & Evolutionary Biology > U. of Tennessee, Knoxville > http://www.brianomeara.info > > Postdoc collaborators wanted: http://nimbios.org/postdocs/ > Calendar: http://www.brianomeara.info/calendars/omeara > > On Fri, Apr 24, 2015 at 3:14 PM, Milton Tan <[email protected]> > wrote: > >> Hello all, >> >> I have a question that is perhaps esoteric, since it's on a method I don't >> see used often. I am looking at the dynamics of body size evolution, and >> have come upon ancestor-vs-change plots described in Alroy 2000 >> ("Understanding the dynamics of evolutionary trends", Paleobiology). This >> is interesting because it will allow me to see if rate of body size change >> depends on body size. I haven't seen this method widely used, so for anyone >> unaware how this works: for each branch, you plot the ancestral state vs. >> the amount/rate of change along the branch. In theory, by looking at a >> scatterplot of ancestral size vs. change, I can answer questions such as >> "Are smaller taxa more likely to evolve to a larger size?" I don't see too >> many people using this method, so I thought I'd ask why. Is there a >> particular reason for it not being used? Are there more powerful methods to >> answer this same question that have supplanted it? >> >> I also have a more specific question, assuming that ancestor-vs-change >> plots are valid. For my dataset, I have reconstructed ancestral states for >> my clade from tip data and pic, and then plotted ancestral states versus >> the rate of change along every branch (attached). While there are a couple >> outliers, you can see the distribution roughly hangs around 0 along the >> entire graph, but I'd like to fit a line so I can present if the slope >> significantly differs from 0 (ie. body size evolutionary rate truly does >> not depend on body size of ancestors). Alroy (2000) describes that the >> shape of the plot can be informative on different dynamics of evolutionary >> change, implying fitting of linear and polynomial lines, but doesn't really >> discuss how to test this statistically. Alroy seems to use this method and >> fits a polynomial line in Alroy 1998 ("Cope's Rule and the Dynamics of Body >> Mass Evolution in North American Fossil Mammals", Science), but he doesn't >> really describe his statistics ! >> in depth. >> >> So my question is how can I fit a line to this data? Problematically, >> since each branch has a sister branch that shares the same ancestral node, >> each ancestral state is represented twice. I think this makes the >> observations non-independent. If they are, this makes me think that a >> linear regression is inappropriate, but I'm not sure. I could average the >> amount of change along each branch for every ancestral node, but I'm not >> sure if that's the best way. Does anyone have any insight on an appropriate >> way to determine a best fit line statistically? >> >> Thanks in advance, >> >> Milton Tan >> Auburn University >> Department of Biological Sciences >> PhD Candidate >> >> [[alternative HTML version deleted]] >> >> _______________________________________________ >> R-sig-phylo mailing list - [email protected] >> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo >> Searchable archive at >> http://www.mail-archive.com/[email protected]/ >> > > [[alternative HTML version deleted]] > > _______________________________________________ > R-sig-phylo mailing list - [email protected] > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/[email protected]/ -- David W. Bapst, PhD Adjunct Asst. Professor, Geology and Geol. Eng. South Dakota School of Mines and Technology 501 E. St. Joseph Rapid City, SD 57701 http://webpages.sdsmt.edu/~dbapst/ http://cran.r-project.org/web/packages/paleotree/index.html _______________________________________________ R-sig-phylo mailing list - [email protected] https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/[email protected]/
