Anatomical partitioning has little influence in topologies from Bayesian phylogenetic analyses of morphological data

Morphological data is a fundamental source of evidence to reconstruct the
Tree of Life, and Bayesian phylogenetic methods are increasingly being
used for this task, along with, or instead of, traditional parsimony
approaches. Bayesian phylogenetic analyses require the use of proper
evolutionary models and their performance have been intensively studied in
the past few years, with significant improvements to our knowledge
regarding their performance. Notwithstanding, it was only recently that
partitioned models for morphology received attention in studies of
empirical data, but a systematic evaluation of its performances using
simulations was never performed. Here we evaluate the influence of
partitioned models defined by anatomical criterion in the precision and
accuracy of consensus tree topologies, evaluating the possible negative
effects of under and overpartitioning. For that, we analysed datasets
simulated using parameters and properties of two empirical datasets, using
Bayesian phylogenetic analyses in MrBayes. Additionally, we reanalysed 32
empirical datasets for diverse groups of vertebrates, applying
unpartitioned and partitioned models. We found that in general,
partitioning by anatomy has little to no influences in the performance of
Bayesian phylogenetic methods in respect to the metrics studied here, with
analyses under alternative partitioning schemes presenting very similar
tree precision and accuracy. We discuss the possible reasons for the
disagreement between the results obtained here and previous studies for
empirical morphological data, and with empirical and simulation studies of
molecular data, discussing the adequacy of anatomical partitioning
relative to alternative methods to partition morphological datasets and
how morphological and molecular partitioning are related.