Repeat after me: evolutionary rates are not the same as mutation rates.https://twitter.com/CEPIvaccines/status/987299962575781888 …
That PNAS paper is problematic IMO. They also used the wrong sequences from the beginning of the outbreak in Guinea - those sequences were corrected a long time ago (Günther’s group) and pushed rate from 1.5 > 2.0. Main issue not model - it’s data.
-
-
Yes, and structured models could easily fix the contrast in sampling intensity between Guinea and Sierra Leone, but it's too easy to say in retrospect. At the time I doubt reviewers would have accepted a deme with 3 sequences without careful sensitivity analysis to priors.
Bedankt, Twitter gebruikt dit om je tijdlijn te verbeteren. Ongedaan makenOngedaan maken
-
-
-
There is also the issue that the structured coalescent over-estimates the time to the root by a factor of 2. That is an important parameter to get correct (and has an interpretable meaning).
-
The path between the three Guinea sequences includes the root of the entire tree, so they end up getting the "correct" final root height from all sequences, which is ~ Dec 2013. I'll assume it's the correct answer for the wrong reasons until I see a prior sensitivity analysis.
- Nog 1 antwoord
Nieuw gesprek -
-
-
Yeah - what he said.
Bedankt, Twitter gebruikt dit om je tijdlijn te verbeteren. Ongedaan makenOngedaan maken
-
-
-
The paper’s main point derives from simulated seqs, not empirical ones, and is well made: simple tree priors can affect estimates when tree is from more complex model.
@evogytis is right that structCoal is one option... -
...and developing a greater range of tree priors is important - we’ve been working on a multifurcating prior. The Discussion has some good insights (incl. the difference between tree height and length inference).
Einde van gesprek
Nieuw gesprek -
Het laden lijkt wat langer te duren.
Twitter is mogelijk overbelast of ondervindt een tijdelijke onderbreking. Probeer het opnieuw of bekijk de Twitter-status voor meer informatie.