We have been walking through McLain et al (2018) and their attempt to put feathers on dinosaurs using statistical baraminology. So far, we have made it to figures 30 and 31. In previous parts, we’ve seen some indeterminate results, as well as some baraminologically untenable results. In this part, we will look at a very small subset of the data, in imitation of McLain et al. In this one, they broke the Brusatte dataset down to just two groups: Therizinosaurs and Oviraptorosaurs. In this group there were just ten taxa. Only 68 characters were used for analysis out of 860.
The first thing to examine is the bootstrapping values as McLain et al, in defiance of traditional statistical baraminology, did not report bootstrapping values. BDISTMDS showed bootstrapping vlaues that varied from very good (100%), to very poor (44%). About half were good, the other half were fair to poor. The most common values was in the 70s, with the second most common being 100. BARCLAY produced similar numbers. While this is not the worst bootstrapping values we have seen, they do not meet the arbitrary cut off to be good in many cases (90%).
Having established the bootstrapping values, we will look at the outputs for these few taxa. Interestingly, even though I used the same taxa, I got different results than the McLain et al paper using BDISTMDS.
I’m unsure why my results for Therizinosaurs and Oviraptorosaurs differ from McLain but they do. McLain et al found some additional continuity between Avimimus and Chirostenotes and additional discontinuity between Erlikosaurus and Citipati and Chirostenotes. It is unclear why this difference exists. We used the same number of characters, and the same taxa. Theoretically this should produce the same result, when run on the same program, but it did not. Was a modification made to BDISTMDS (excluding for taxic relevance which is immaterial in this instance as the same number of taxa are used for both)? If so, why was it not reported? Is it possible that changing the taxic relevance to be more broad also changes the way the BDC calculations are done? We don’t know. Dr. Wood has been very reticent to share his code with people who do not agree with him. The best that is available is Dr. Cserhati’s attempt to duplicate it. In any case, as expected, the BARCLAY results differ from both BDISTMDS results.
As you can see the results for BARCLAY are substantially the same with slight differences. Discontinuity and continuity are detected in new places that BDISTMDS did not detect them. The blocks of taxa are substantially the same, though Incisivosaurus links with the larger block in this scenario, where it did not previously. Again, BARCLAY is finding continuity where BDISTMDS did not. Is BARCLAY going to be across the board worse at finding discontinuity than the already failing grade (55%) that BDISTMDS has? It’s beginning to look that way.
Interestingly, I ran the BDISTMDS and BARCLAY at 85% and produced an excellent set of graphs as shown below. While only 12 characters are used, I find it fascinating how clear cut the groups are.
As you can see, the BARCLAY detects significantly more discontinuity and slightly more continuity than BDISTMDS. Given how small the sample size it, I don’t want to draw conclusions from that, but it is very interesting that it is a much clearer picture with a higher relevance cut off, highlighting the insanity of using 75% as the standard.
Having presented the graphs, I then obtained MDS results. As you can see, the 85% showed one tight cluster and the other spread out while the 75% showed everything spread out.
What can we make out of all this? Well, the Therizinosaurs and Oviraptorosaurs are probably separate kinds. But that was something we probably could have determined intuitively without appealing to statistical baraminology. So far as I can tell under this system, there is nothing egregiously wrong with these results, but, given how unreliable other subsets of the data are, it would be wise to keep an open mind on these results as we move further through the paper.
Do you know what’s going to happen when you die? Are you completely sure? If you aren’t, please read this or listen to this. You can know where you will spend eternity. If you have questions, please feel free to contact us, we’d love to talk to you.
In His Image 