Baraminic Distance, Bootstraps, and BDISTMDS
Todd Charles Wood
Occas. Papers of the BSG No. 12, pp. 1-17
©2008 BSG.
|
Printable PDF of this article (1.47 Mb)
OPBSG Home
BSG Home
|
| Abstract |
Although baraminic distance correlation (BDC) has been used to define more holobaramins than any other method, the
technique remains largely untested. One method of testing the reliability of BDC conclusions is bootstrapping. A bootstrapping
procedure has been added to the software package BDISTMDS (http://www.bryancore.org) that randomly resamples the data matrix
100 times and which calculates baraminic distances, correlations, and probabilities for all taxon pairs in each resampled dataset. Applied
to BDC, bootstrapping can reveal which correlations between taxa are robust to random perturbations in the character data and which are
dependent on the particular character states of the native dataset. Bootstrap values were calculated for ten previously-published BDC
analyses. Results indicate very good bootstrap values for the majority of taxon pairs (median 100%; 70.2% of taxon pairs had bootstrap
values >90%). Bootstrap values are proportional to the number of taxa in the dataset but are not correlated with baraminic distances.
As a method for evaluating the robustness of baraminic inferences from BDC, bootstrapping should be used in future baraminology
research.
|
|