Mega-tree construction guidelines

Since Peter Stevens has put his (dynamic) hypothesis for the phylogenetic relationship among the APG orders (on APweb), we have been using this as the backbone of our mega-tree, and are extremely grateful for his effort. We had previously been compiling our tree from the most recent 'whole angiosperm' study, e.g. Soltis et al. (2000). Peter has extensively documented his sources for tree construction down to the family level.

Our 'resolved' trees (name prefixed by 'R') use the complete resolution that Peter has decided on on APweb. Our 'conservative' trees (name prefixed by 'C') remove any branches that Peter has annotated as having bootstrap support less that 80%, or when a whole tree is indicated as having 'weak support.'

To the APweb tree, we are adding family-level phylogenies from published studies. Our method for deciding which tree to include is:

1. Where a single most parsimonious tree is presented, this topology is used for our 'resolved' tree. The 'conservative' tree is based on the strict consensus tree given, with branches of support less than 80% removed.
2. Where parsimony and maximum likelihood analyses are both presented, the former result is used, not as a statement about the relative merit of the two approaches, but for consistency.
3. The results from a more focused study (on particular nodes at any taxonomic level) over-rides conflicting results of a broader study (for the ingroup only).
4. Studies using a greater number of informative characters are given dominance over studies with fewer characters.
5. Where no molecular or morphological phylogenetic analyses have been published, sub-familial classification and authors' published 'intuitions' about relationships are used in the 'resolved' tree.