Just thinking out loud here, but maybe one could use a rate matrix that doesn't allow gains of 1, and instead reroot the tree at various points with a fixed root state of 1? The only semi tricky bit would be searching for the optimal position of this 'root', which is where the single gain of 1 occurs...
I was about to hit send on this suggestion and then realized it ignores that the rootward remainder below the artificial 'root' also has to be constrained to be zero, which seems even trickier. Hrm. -Dave On Monday, September 14, 2015, Graeme Lloyd <graemetll...@gmail.com> wrote: > Hi Everyone, > > Thanks for your responses. I guess I failed to explain what I really > meant. Apologies. > > I have a time-scaled ultrametric tree and a binary character (and no > missing data) and I want to fit a true Dollo character (exactly as Brian > describes with: 1) a root state of zero, 2) a single gain, and 3) losses if > required). The thing I want to get at is those losses, and ideally using > stochastic character mapping so I have both where and when and > uncertainties in both. > > If I understand correctly the matrix(c(0,0,1,0),2,2) model won’t quite do > this so maybe ace/make.simmap can’t cope with this? > > Thanks again, > > > Graeme > > > On 15 Sep 2015, at 10:36, Liam J. Revell <liam.rev...@umb.edu > <javascript:;>> wrote: > > > > Hi Graeme. > > > > What Matt says is correct, of course, but it is also true that you can > do this with ace or make.simmap. > > > > That would look like: > > > > model<-matrix(c(0,0,1,0),2,2) > > rownames(model)<-colnames(model)<-0:1 > > ## ace > > fit<-ace(x,tree,type="discrete",model=model) > > ## make.simmap > > trees<-make.simmap(tree,x,model=model,nsim=100) > > > > You (should) get the exact same parameter estimates for the rate 1->0 > for all three methods, so it depends what you want to do: stochastic > mapping, marginal ASR, joint ASR, or simply estimate the parameters of the > Mk model. > > > > BUT (and I see that Brian just responded with something along the same > lines) a Dollo character does not imply that 0->1 does not occur, but that > it occurs only once. If the root state is 0, not 1, then this could make > things a little more complicated; however one possibility (and, to be fair, > I haven't given this too much thought), would be to put a strong prior on > the root to be 0 and the transition 0->1 to be low (but non-zero) using > stochastic mapping. > > > > All the best, Liam > > > > Liam J. Revell, Associate Professor of Biology > > University of Massachusetts Boston > > web: http://faculty.umb.edu/liam.revell/ > > email: liam.rev...@umb.edu <javascript:;> > > blog: http://blog.phytools.org > > > > On 9/14/2015 8:15 PM, Matt Pennell wrote: > >> ## make a likelihood function > >> lik <- make.mk2(t, d) > >> > >> ## constrain the likelihood so that it only allows forward transitions > >> lik_con <- constrain(lik, q10~0) > >> > >> ## estimate the MLE > >> fit <- find.mle(lik_con, 0.1) > >> > >> ## use parameters to do ASR using ML parameters. Other operations also > >> possible. > >> asr <- asr.joint(lik_con, fit$par) > > _______________________________________________ > R-sig-phylo mailing list - R-sig-phylo@r-project.org <javascript:;> > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at > http://www.mail-archive.com/r-sig-phylo@r-project.org/ > -- 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 [[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/
