Yeah, I was concerned it might have been related to a zero-branch-length problem, but some simulation shows that only Graham's fix will let fitContinuous run with a simulated paleo-tree. -Dave
library(paleotree) library(geiger) taxa<-simFossilTaxa_SRCond(avgtaxa=30,p=0.1,q=0.1,r=0.1) rangesCont<-sampleRanges(taxa,r=0.1) cladogram<-taxa2cladogram(taxa) #get the true tree at the first appearance times trueTree<-taxa2phylo(taxa,obs_time=rangesCont[,1]) trait<-rTraitCont(trueTree) #no solution fitContinuous(trueTree,trait,model="EB") #nope #my solution fitContinuous(addTermBranchLength(trueTree),trait,model="EB") #nope #Graham's solution fitContinuous(trueTree,trait,model="EB",bounds=list(alpha=c(0,1))) #yep On Sun, Nov 4, 2012 at 1:17 PM, Frank Burbrink <frank.burbr...@csi.cuny.edu>wrote: > Just to confirm that Graham's fix works. > > Using the following libraries (and their dependencies) I conducted the > simulations: > > library(TreeSim) > library(phytools) > sim.bd.taxa(30,1,lambda=0.1,mu=0.05)[[1]][[1]]->tree > fastBM(tree)->data > > fitContinuous(tree,data,model="EB") > Fitting EB model: > Error in solve.default(phyvcv) : > Lapack routine dgesv: system is exactly singular > ###This now works with a parameterized alpha (below)...not sure how > good the estimates are (they were simulated under ###BM after all) > > fitContinuous(tree,data,model="EB", bounds=list(alpha=c(0,1))) > Fitting EB model: > $Trait1 > $Trait1$lnl > [1] -264.4297 > > $Trait1$beta > [1] 0.8355526 > > $Trait1$a > [1] 0.001813503 > > $Trait1$aic > [1] 534.8593 > > $Trait1$aicc > [1] 535.1042 > > $Trait1$k > [1] 3 > > On Sun, Nov 4, 2012 at 2:01 PM, David Bapst <dwba...@uchicago.edu> wrote: > > Nicolas, > > Not certain how you are time-scaling your tree, but I often get warnings > of > > singularity from various comparative analysis functions due to the > presence > > of very short terminal branches. I have found that such warnings of > > singularity when dealing with paleo-trees can often be avoided by adding > a > > very small amount to all the terminal branch lengths, say 0.001. This is > > very simple to do with a few lines of code; alternatively, the function > > addTermBranchLength() in the library paleotree does this automatically > if a > > tree is passed to it. > > > > Also, I would advice you to consider multiple methods for time-scaling > your > > tree of fossil taxa and compare results across those trees, if you are > not > > doing so already. In practice, I find one can sometimes give very > different > > results for some comparative analyses with different time-scaling methods > > on the same paleontological dataset. > > -Dave > > > > > > On Sun, Nov 4, 2012 at 10:59 AM, Nicolas Campione < > > nicolas.campi...@mail.utoronto.ca> wrote: > > > >> Dear R Phylo List, > >> > >> I'm trying to test an Early Burst model of evolution against a tree > using > >> the 'fitContinuous' function in 'geiger'. However, I get the following > >> error message, and am unsure as to what I can do to fix this: > >> > >> > fitContinuous(tree,trait,model="EB") > >> > >> Fitting EB model: > >> Error in solve.default(phyvcv) : > >> Lapack routine dgesv: system is exactly singular: U[4,4] = 0 > >> > >> My tree is based on fossil taxa and my branch lengths are calculated > using > >> time (in millions of years). I seem to be able to test every other > model, > >> only EB is giving me troubles. I'm pretty sure it has something to do > with > >> the branch lengths, cause if I used other measures, such as all=1 or > >> grafen, I don't get an error. > >> > >> Has anybody had this issue before? > >> > >> Any assistance would be greatly appreciated. > >> > >> Cheers, > >> > >> -Nic- > >> > >> ---- > >> Nicolás E. Campione > >> M.Sc., Ph.D. Candidate > >> Dept. Ecology & Evolutionary Biology > >> University of Toronto > >> 25 Wilcocks St. > >> Toronto, ON > >> Canada M5S 3B2 > >> > >> Royal Ontario Museum > >> 100 Queen's Park > >> Toronto, ON > >> Canada M5S 2C6 > >> Office: 416-586-5591 > >> Email: nicolas.campi...@mail.utoronto.ca > >> > >> [[alternative HTML version deleted]] > >> > >> > >> _______________________________________________ > >> R-sig-phylo mailing list > >> R-sig-phylo@r-project.org > >> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > >> > >> > > > > > > -- > > David Bapst > > Dept of Geophysical Sciences > > University of Chicago > > 5734 S. Ellis > > Chicago, IL 60637 > > http://home.uchicago.edu/~dwbapst/ > > http://cran.r-project.org/web/packages/paleotree/index.html > > > > <http://home.uchicago.edu/%7Edwbapst/> > > > > [[alternative HTML version deleted]] > > > > > > _______________________________________________ > > R-sig-phylo mailing list > > R-sig-phylo@r-project.org > > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > > > > > > -- > NEW EMAIL ADDRESS: > frank.burbr...@csi.cuny.edu > *********************************** > Frank T. Burbrink, Ph.D. > Professor > Biology Department > 6S-143 > 2800 Victory Blvd. > College of Staten Island/CUNY > Staten Island, New York 10314 > E-Mail:frank.burbr...@csi.cuny.edu > Phone:718-982-3961 > Web Page: http://scholar.library.csi.cuny.edu/~fburbrink/ > *********************************** > > ________________________________ > > Washington Monthly< > http://www.washingtonmonthly.com/magazine/septemberoctober_2012/features/americas_bestbangforthebuck_co039461.php> > magazine ranks the College of Staten Island as one of Americas > Best-Bang-for-the-Buck Colleges > -- David Bapst Dept of Geophysical Sciences University of Chicago 5734 S. Ellis Chicago, IL 60637 http://home.uchicago.edu/~dwbapst/ http://cran.r-project.org/web/packages/paleotree/index.html <http://home.uchicago.edu/%7Edwbapst/> [[alternative HTML version deleted]]
_______________________________________________ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo