AWTY and Tracer assess convergence of different results of the analysis and should both be used. AWTY will help you determine if your tree topologies are converging on the same space. Tracer will help you determine if the parameters of your models have converged. Often you only see mention of assessing convergence in Tracer, but convergence of parameters does not guarantee convergence in topologies.

In general, in Tracer you want ESS values greater than 200 for all your parameters. When looking at MrBayes analyses you want to be sure that the ESS values are above 200 for the combined parameter files. This ensures good mixing and convergence.

For AWTY you want to upload your tree files. I find the most useful diagnostic for most phylogenetic purposes is the compare function. This compares the bipartition frequency between multiple runs. If the runs converge, they should line up on the diagonal. Any deviation from the diagonal is the difference in support between the two runs. These differences will reduce your node support value in the consensus tree. A small amount of deviation from the diagonal is probably fine, but larger differences indicate lack of convergence in tree topology and analyses should be redone.

They have a pretty good tutorial: http://beast.bio.ed.ac.uk/Analysing_BEAST_output. ESS values might give you a clue in this regard

Hello, Hazel!

A have little experience with tracer and awty.

Recommend you to check out tutorials

here

https://molevol.mbl.edu/wiki/index.php/MrBayes_lab#Using_other_programs_to_summarize_MCMC_results

and here

http://treethinkers.org/tutorials/mcmc-diagnosis/

You should inspect log-likelihood curve and split frequency.

Great! Many thanks all, I will check these out today!

AWTY and Tracer assess convergence of different results of the analysis and should both be used. AWTY will help you determine if your tree topologies are converging on the same space. Tracer will help you determine if the parameters of your models have converged. Often you only see mention of assessing convergence in Tracer, but convergence of parameters does not guarantee convergence in topologies.

In general, in Tracer you want ESS values greater than 200 for all your parameters. When looking at MrBayes analyses you want to be sure that the ESS values are above 200 for the combined parameter files. This ensures good mixing and convergence.

For AWTY you want to upload your tree files. I find the most useful diagnostic for most phylogenetic purposes is the compare function. This compares the bipartition frequency between multiple runs. If the runs converge, they should line up on the diagonal. Any deviation from the diagonal is the difference in support between the two runs. These differences will reduce your node support value in the consensus tree. A small amount of deviation from the diagonal is probably fine, but larger differences indicate lack of convergence in tree topology and analyses should be redone.

Hi Charles , this is very useful, many thanks. It's certainly helpful to understand what values i should be looking for in these programs, and did not understand fully what each program offered. I have been trying it upload my tree files into AWTY but have had no luck so far as it keeps freezing. I'll figure it out I'm sure! Thanks again!

Uploading the files to the AWTY web server can be difficult. I find uploading while on a University campus with faster internet helps, I have a hard time getting it to work at home. There is a size limit to the files that can be uploaded, so check that you aren't exceeding that. If your files are too big, or if you just want to reduce the transfer time, you can reduce the frequency of your sample of the posterior. This can either be done when the analysis is run or by using logcombiner from the BEAST software package. Logcombiner can resample your tree files at a lower frequency to reduce file size which may help with the transfer. You really only need a thousand or so samples from the posterior. There are direction on the BEAST website for logcombiner.

Good idea! I'll try it on campus tomorrow. I also look at reducing the number of trees as you suggest, I've used logcombiner before so i'm fairly familiar with it (even tho it was some time ago). Thanks again, VERY helpful!