I would have to see your code to know 100% for certain - but your results make 
intuitive sense, so let me just guess. When you have OU under a very high alpha 
(which is what you are getting here), then you are modeling a process following 
a random walk with an extremely high tendency to be pulled towards the OU 
"optimum" value. In this case, characters do not retain any information about 
their past states, and so the "best guess" for every node in the tree is the 
mean of all of the characters. You can check to see if the value you are 
getting back is actually the mean of all of your tips.

To me, what this means is that if you think the model is correct (lambda or OU) 
then you have no information about any of the ancestral character states in the 
tree.

lh

On Nov 8, 2012, at 3:14 PM, Zachary Chejanovski <[email protected]> 
wrote:

> Hi,
> 
> I am attempting to estimate ancestral states using the following methods-
> APE::ace, GEIGER::getancstates, as well as this method recommend by Liam 
> Revell for reconstructing under an OU model:
> 
> ou<-fitContinuous(tree,data,model="OU")$Trait1 
> anc.mins<-vector()
> N<-length(tree$tip)
> M<-tree$Nnode
> outree<-ouTree(tree,ou$alpha)
> for(i in 1:M+N){
>       
> anc.mins[i-N]<-ace(data,multi2di(root(outree,node=i)),method="pic")$ace[1]
>       names(anc.mins)[i-N]<-i
> }
> Here are the parameter estimates for one of my traits:
> $Lambda
>        lnl     beta    lambda      aic     aicc k
> 1 -121.4771 31.33674 0.4057359 248.9541 250.2173 3
> $OU
>        lnl     beta      alpha convergence                                    
>      message k      aic     aicc
> 1 -120.7098 281.9509 0.08288029           0 CONVERGENCE: REL_REDUCTION_OF_F 
> <= FACTR*EPSMCH 3 247.4196 248.6828
> 
> When I do not transform the tree (i.e. Brownian Motion), the values returned 
> seem valid. But when I transform the tree using either lambdaTree, ouTree. or 
> even the exponentialTree, all the values returned are almost identical and 
> when I plot them on the phylogeny, it is evident that they do not make sense 
> in regards to the tip values.
> 
> Any help on this would be greatly appreciated.
> 
> Thank you,
> Zac
>       [[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-phylo mailing list
> [email protected]
> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo

Luke Harmon
Associate Professor
Biological Sciences
University of Idaho
208-885-0346
[email protected]

_______________________________________________
R-sig-phylo mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo

Reply via email to