To investigate major evolutionary trends and the importance of fossil data weneed to be confident that both phylogenetic trees and fossil dates are reliable. Indicesof stratigraphic congruence provide a way to quantify the fit between the fossil recordand phylogeny, but appear to be subject to a number of putative biases. I used bothsimulated data and a large sample of empirical trees to determine the effect of thesebiases on the most widely used indices of stratigraphic congruence to determine. TheGER* (the modified gap excess ratio) was the least sensitive and thereforerecommended for use. I found that stratigraphic congruence varied significantly acrosshigher taxa (for example, arthropods were less stratigraphically congruent thantetrapods), and also throughout the Phanerozoic (the last 540 million years), closelyfollowing the taxonomic composition of my sample.I focussed on data quality and in particular taxon sampling, homoplasy and treesupport to investigate general trends across taxonomic groups. A novel script wasdeveloped to automatically carry out continuous taxon jackknifing to investigate theeffect of taxon sampling on the stability of phylogenetic trees. While this is acomputationally intensive process, I found that measures of homoplasy and support(which are much easier to calculate) could serve as partial indicators of whether a treewas likely to be sensitive to taxon sampling. There was no major variation in taxonsampling trends across higher taxa. A modified version of this script was then used tolook at particular cases of conflicting phylogenetic hypotheses to determine how easyit would be to get a data set to generate a constraint topology with only small changesto the taxon sample. In almost every case, it required maximal removal of taxa fromthe data set in order to match the constraint topology, indicating that there were verydifferent phylogenetic signals in the different data sets.The extent of trends across taxonomic groups and through time is variable.Although stratigraphic congruence varies significantly between groups and throughoutthe Phanerozoic, measures of homoplasy and support do not appear to be taxondependent. Taxon sampling is an important consideration when designingphylogenetic analyses: denser taxon sampling can have a positive influence onestimates of phylogenetic accuracy and perturbations of the taxon sample can result in radically different evolutionary relationships.
|Date of Award||6 Mar 2014|
|Supervisor||Matthew Wills (Supervisor)|