Resampling
From TNT
Resampling supports can be based on different types of character. Standard bootstrapping is influenced by uninformative characters (and by characters irrelevant to monophyly of a given group). Bootstrapping (both standard and Poisson) and jacknifing (except for a resampling probability of 50%) are affected by character weight and transformation costs (e.g. additive characters). Symmetric resampling is affected by none.
Results can be output using frequency, frequency difference (GC: Group present / contradicted) and frequency slopes (Goloboff et al., 2003). Outputting results with the frequency may produce alterations in the apparent support for groups with very low support (and it cannot measure support for groups with very low support). Both GC and frequency slopes solve this problem (for slopes, supported groups have negative slopes; the closer to 0, the better supported the group).
The supports can be measured for groups present in any given tree, with or without a particular cut-off value to collapse nodes with a low support. Normally, they will be shown for the most parsimonious tree, but other trees can be used to measure how strongly contradicted some groups are. The slopes must be measured by reference to a pre-existing tree. TNT does not present the supports in the shape of a majority consensus tree.
If some taxon has its position in the tree very poorly defined, it may strongly decrease the support for many groups in the tree. If you wish to find out whether the rest of the tree is well supported, you can eliminate the taxon from the consensus when resampling. To find out which taxon is floating around you may have to first do a more superficial estimation, saving the trees, and trying to prune different taxa (note that in such case you have to take the precaution of not collapsing the trees with SPR or TBR, as otherwise you may never find the floating taxon: once you found it, you turn collapsing back to SPR or TBR –the most effective—and do a more careful estimation of the resampling frequencies eliminating the evil terminals).
Resampling is invoked with the command resample.
Path:Analyze/Resampling
Plotting resampling values onto a tree
Suppose you already have your strict consensus and you are interested in plotting the values of jacknifing for the nodes in that tree.
- Be sure that the consensus tree is in the memory.
- Turn on the tree tags storing. In the menu version go to Tree→Multiple tags→Store tree tags. The command to do that is
ttags = - Perform the Resampling analysis.
- To print the tags on the tree go to Trees→Multiple tags→Show/Save tags. The tree with the labels can be saved as a graphic pressing the letter M. The command to do that is
ttags ;
Plotting bremer support values and resampling values on the same tree is very similar. The only thing you should take care of is that when doing your bremer analysis you will have lots of suboptimal trees held in memory but not the consensus where you want to plot your values. Hence you have to clear the tree buffer after running your bremer (keep 0) and then open the file where you have already stored your consensus. After that you can go to Tree/Multiple tags/show-save tree tags to see both legends: bremer support and resampling.
