Some comments:
1) In the "Epidemiology model", we always set bdss.r = 1 in our analyses (as
performance was bad for r<1). Now in Beauty by default there's some prior on
[0,1] - shouldn't we have the default as r=1?
And shouldn't we set bdss.sampleProbability=0 as we do not sample extant
species in epidemiology..
As psi is correlated with birth and death (ie the three are non-identifiable),
we should put a strong prior on one of the rates (eg let the user specify
psi/mu)?
Stadler et al (2011) : Estimating the basic reproductive number from viral
sequence data, Mol.Biol.Evol., revision submitted, July 2011.
is meanwhile published:
Stadler et al (2011) : Estimating the basic reproductive number from viral
sequence data, Mol.Biol.Evol., doi: 10.1093/molbev/msr217, 2011.
2) In the "speciation: birth-death serially sampled model": it says in the
description "Yule Process with serial samples" - but we should write
birth-death instead of Yule!
As psi and rho correlate with birth and death rates (ie are non-identifiable),
we should put by default very strong priors on them - eg let the user specify
rho and psi/mu?
3) In the "speciation: birth-death process": delete "in press" in the citation
One very general comment: How did Denise now put priors for the BDskyline model?
Eventually we should probably use the same parameterizations as Denise, since
the simple birth-death serially sampled model has the same pareameter
correlations as Denise's skyline model (and I think Denise used psi/mu that's
why I suggested it above)!
Cheers Tanja
Original issue reported on code.google.com by dong.w.xie@gmail.com on 5 Dec 2011 at 2:18
Original issue reported on code.google.com by
dong.w.xie@gmail.com
on 5 Dec 2011 at 2:18