In part one, we looked at the first dataset used in McLain et al’s attempt to argue that dinosaurs had feathers. In this post, we will look at some subsets of the data that McLain et al used, singling out the Paravian members of the dataset. The Paraves group includes modern birds, extinct birds like Archaeopteryx, as well as dromaeosaurs, troodons, and other extinct animals typically classified as dinosaurs.
My first step was to recreate the dataset McLain et al used. This done, I ran the dataset through both BDISTMDS and BARCLAY. Of the 860 characters in the dataset, only 100 met the 75% relevance cut off to be used. Bootstrapping results, which, as a reminder, are meant to tell you how good the data is, were good in some cases, but in many cases were sub-80%. 95% or more is considered a good bootstrap value. The lowest value I noticed in my scan through the numbers was 46, and a significant portion of the bootstrapping numbers were in the fifty to seventy percent range. This does not inspire confidence in the data. Again I have to question why McLain et al never addressed bootstrapping values in their paper.
Having looked into the bootstrapping values, I wanted to see how much like the McLain et al results mine were. Unsurprisingly, they were similar. BDISTMDS came up with identical results to the McLain paper. Not surprising. What was surprising, was the difference between the BDISTMDS and BARCLAY results. There was significantly less discontinuity in the BARCLAY dataset and a few more cases of continuity. Remember, this is the exact same data, but the two methods are getting different results. It is particularly concerning that BARCLAY is detecting even less discontinuity than BDISTMDS, since, as Wood acknowledged, BDISTMDS only accurately detected discontinuity 45% of the time! Is BARCLAY going to be even worse?
The next step I attempted was to use the 95% relevance cut off. It would not work. However, 85% would work. It cut the number of characters to a mere 19 of an original 860 characters. That tells me one of two things. Either, the taxa under study are so fragmentary that they share very few characteristics, or there are multiple different, unrelated groups under study. I don’t think any creationists, even Mclain et al, would argue that modern birds, dromaeosaurs, and troodontids are related. Given that we are also working with fossils, it is likely there is some fragmentation as well. However this raises some questions.
The first question it raises is why are organisms no creationists believes are related being shown as related? Both BDISTMDS and BARCLAY show links between the raptors, modern birds, dromeosaurs and troodontids. Obviously we have not looked at MDS yet, but MDS is completely subjective. You can look at an MDS plot and make it say anything you want it to say. So this is at least concerning.
Second, and probably a larger issue, is the fact that the dataset starts with so many different kinds and so much fragmentation. This being the case, I have to ask the question of whether anything can be gleaned from this dataset at all, given its fragmented state, and the wide variety of taxa. I suspect this dataset would need to be trimmed down much more, with only related taxa and some outgroups for it to be useful.
Moving on to the MDS, both BARCLAY and BDISTMDS produce very similar MDS plots. They don’t help clarify the issue very much.
Note that in both results, what you have is a cloud with no obvious groupings. This differs from the results presented by McLain et al where they found a four member group of birds. However, what this illustrates is how subjective MDS is. Groupings are entirely dependent on how you choose to rotate the 3D structure and present it. I could argue based on orientation that there is a five member group in the BDISTMDS data on the right edge, but no grouping at all in the BARCLAY data. I’m not going to because they are at different orientations currently, but if I wanted to make the two methods disagree with one another, I could. I’m not saying that’s what McLain et al are doing here, but have no standard, even an arbitrary one, for how to present the data, really weakens any usefulness of MDS.
Taken together, this Paraves section of the data makes the case for the common ancestry of dromaeosaurs, troodons, modern birds, and extinct birds. Obviously I do not believe that, and I don’t think McLain et all do either. BDISTMDS and BARCLAY however, so far, would indicate that they are all members of the same baramin, or that the dataset is unable to distinguish between the groups. Given how much data is not used, I’m not surprised the dataset cannot distinguish. It’s far too fragmentary to use except in subsets. Even the subset taxa must be carefully chosen, and likely consist of organisms suspected to be of the same kind, with appropriate outgroups. We will continue with our analysis of this dataset in an upcoming post.
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.