Estimating error around dates?

[burbrink]

Estimating error around dates?

[josephwb]

FWIW I believe the only route available at the moment is with the tree distribution. This typically involves bootstrap trees, because if you were doing Bayesian analyses you would be using BEAST in the first place!

[burbrink]

Thanks! I figured that was the option.

I suppose I will loop it over the distributions of BS trees then. I guess I could get a Pp distribution on a fixed topology If had to (though for 732 taxa and over 400 loci is not optimal for something like that).

[josephwb]

The tricky bit is to make sure that your constraints (i.e. MRCA statements) are placed consistently on the various trees (I'm sure you know this). In the past I had R code to analyze individual trees and write out individual treePL config files. This would, for example, get rid of redundant constraints. Not too difficult to do, but I can (try to!) find the code and put it in the repo if that would be useful.

[burbrink]

That is for sure an issue.

However, the problem is that with a genomic dataset of this size, I am simply optimizing all of the genome data on the fixed species trees to generate branch lengths in terms of substitution probabilities. I can "easily" generate a distribution of substitution rates on this single species tree, but can't really do anything resembling relaxed phylogenetics. Perhaps this is tense phylogenetics?

[josephwb]

Indeed: the past sure is tense.

Good that you don't have to worry about phylo conflict; will things much easier. Good luck!

[burbrink]

Just a follow up. Say you have your bootstrapped trees (conforming to some species tree) in a folder that also includes your control file (control), the treePL executable, and this loop (treepl_loop.R) then just use this R script to run everything (provided that lambda is the same for each replicate):

x<-list.files()
x<-x[-grep("control",x)]
x<-x[-grep("treePL",x)]
x<-x[-grep("treepl_loop.R",x)]

for(i in 1:length(x)){
y<-readLines("control.txt")
y[1]<-paste("treefile =",x[i])
y[15]<-paste("outfile = outPL_",x[i],sep="")
write(y,"control.txt")
system("treePL control.txt",ignore.stdout = T)
print(i)
}

Co-authored with Marcelo Gehara

[josephwb]

@burbrink That could be useful for running the analyses in parallel (but would require altering your R code from a for loop; not too difficult), but treePL handles multiple trees natively. So you could just do:

cat *.tre > all.trees # if the trees are in different files
treePL control_file # where control_file obviously mentions 'all.trees'

[burbrink]

Good point, I could have just done that with an apply fxn and put in parallel. But this is running lightning fast already (with 8 threads).

I had no idea treePL handled these natively. Reinventing the wheel.

Well kids, that is a good loop for controlling stuff from R anyhow...