This is a fascinating paper I read using Bayesian analysis to compare different graph models to genomic data. Intriguingly, it shows that a dependency graph is a much better fit than a tree model. However, a tree model is what we'd expect from common descent with variation. What could make this dependency graph model, instead?


For those who are worried about site cookies, here is the paper hosted on my Box.net account:


  • 1
    $\begingroup$ This paper was discussed by three scientists and the author here: discourse.peacefulscience.org/t/…. Really interesting paper, but Ewert has not yet made his case. $\endgroup$ Jul 22, 2018 at 16:19
  • $\begingroup$ @Dr.S.JoshuaSwamidass Could you summarize the problems with the paper in an answer? As a layperson, the tree model is what I've always thought of when it comes to evolution. It is hard to imagine how common descent could result in anything else. If you can provide explanation as to how a non-tree could emerge from evolution, that would be very helpful. $\endgroup$
    – yters
    Jul 22, 2018 at 23:37

1 Answer 1


It appears the controversy stems from a misunderstanding of what the log Bayes factor in the paper means. Dr. Swamidass states the D. Graph vs Tree result should be divided by the Tree vs Null result to show the D. Graph only explains the data 1.7% better than the Tree.

However, the author claims the log Bayes factor actually means the D. Graph is $2^{111,823}$ times better than the Tree. Which clearly shows common descent (Tree) is a vastly inferior explanation for the data than the D. Graph.

The issue is there is an enormous amount of data, which can artificially inflate the log Bayes factor. However, the proportion should stay stable. Hence the focus on the 1.7%. Explained here:


  • $\begingroup$ Really? You ask a question about a paper that doesn't seem to be in Pubmed, and then you give yourself credit for the best answer? $\endgroup$
    – swbarnes2
    Jul 26, 2018 at 21:51
  • $\begingroup$ No other answer is forthcoming. Do you have one? $\endgroup$
    – yters
    Jul 26, 2018 at 22:29
  • $\begingroup$ No way am I giving a click to let some rando site put cookies on my computer. Grown-ups cite Pubmed, so everyone can see that the paper in question is legitimate. $\endgroup$
    – swbarnes2
    Jul 26, 2018 at 22:57
  • $\begingroup$ I've hosted it on my box account if you would prefer that: app.box.com/s/2mtiwafzr67hsee8o9m7bnfjr7dbmzsv $\endgroup$
    – yters
    Jul 27, 2018 at 12:54
  • $\begingroup$ Kids are curious about the world :) $\endgroup$
    – yters
    Jul 27, 2018 at 13:21

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .