Mr Bayes query - startvals command (Jul/08/2011 )

I'm just starting with baysian analysis using MrBayes. I think my analysis might have trouble converging so I am trying to add a newick tree using a tree block which is recognised by a startval command in the MrBayes block...but when I run the file it cannot execute because it "cannot find command "startval""

Does anyone have any suggests for what might be causing this problem? I've included newick trees with and without branch length, tried spelling variations, varied the position of the command within the block. It is possible that my tree have polytomies... but I've run an alternative file with only bifurcating branches and it has the same problem.

Also, any alternative suggests for helping the converging process along would be appreciated!

Cheers
M

Cannot help you with your problem, I have never tried this myself. But mrBayes is often causing trouble with space, points, etc in the species names and with long, similar names of the sequences, probably this is the same when using a tree so you should try to change the annotation of your tree (or even in the original file/data file of your tree)....

However, usually the strength of mr bayes is that most analyses will converge after some time (often 5-10 million generations needed; running more chains than suggested in the default options) without any prior modifications. Have you tried to run your dataset for so many generations? I know this can be frustrating and time consuming, depending on your dataset but usually it pays off with a reasonable tree that is more or less well supported.

Do not know what kind of dataset you are working with - probably this might justify using pre-defined tree - but I would always try to run un-biased analyses in the first place; and if this does not work out you can try to increase or decrease the heat or change the values for the single bases. I suppose you tested the best possible model for your analysis - so this will give you additional clues which values could be weighted more than others in order to get a well supported tree.

The MrBayes developers started a mailing list for discussing problems:
https://lists.sourceforge.net/lists/listinfo/mrbayes-users

Thanks for the suggestions. I ran the analysis with out the tree block, got to convergence and then came up with an error message 'could not find parametre "relburnin"' despite having sumpt and sump instructions in my mrbayes block. There is a patch for sump error mentioned in the Mr Bayes mailing list so fingers crossed, that might help.