[Taxacom] Dispersal paradigm (was: Antarctic dispersal)

[Richard Zander]

If the prior affects the posterior probability too much, then the exercise fails. The whole idea of Bayesian analysis is to update previous estimates with new information. If there is not much of an update (use Bayes factors) then Bayes is not the way to go. You can use the simple joint probability distribution instead, and maybe a likelihood ratio. Any info on this?

-----Original Message-----
From: taxacom-bounces at mailman.nhm.ku.edu [mailto:taxacom-bounces at mailman.nhm.ku.edu] On Behalf Of John Grehan
Sent: Friday, June 13, 2014 7:55 AM
To: Michael Heads
Cc: taxacom at mailman.nhm.ku.edu
Subject: Re: [Taxacom] Dispersal paradigm (was: Antarctic dispersal)

Ken and Jason,

I was hoping to see either of you explain your positions with respect to the points published by Mike, but Mike has kindly summarize his findings again and so now I look forward to to seeing how you regard the points raised.

John Grehan


On Thu, Jun 12, 2014 at 11:24 PM, Michael Heads <m.j.heads at gmail.com> wrote:

> Hi Ken,
>
> All the Bayesian dating studies I've seen use very tight priors (see 
> my review in J Biogeog 37: 1749. 2011, also my 2014 book). In other 
> words, they stipulate, before any analysis, that the clade age can 
> only be a little older than the oldest fossil of the group (e.g. 2 million years).
> This generates 'evidence' that the group is young, and so vicariance 
> is ruled out and dispersal is 'revealed'. Many studies use exponential 
> priors (which stipulate that the probability of earlier clade ages 
> decreases very rapidly), and cite Ho as support for this (even though 
> Ho specifically warned against this). Many studies stipulate that the 
> clade age can be no older than the base of the geological formation 
> that the oldest fossil is from (typically just a few million years) - 
> again, there is no logical basis for this, but it generates the desired results.
>
> A few authors apart from myself have pointed out the problem of the priors.
> Berbee and Taylor (2010) found that 'Different priors for node ages 
> had a striking effect on our results....?, but ?Whether or not it is 
> possible to select reasonable priors is a serious concern in Bayesian 
> analysis (Felsenstein, 2004). Parham et al. (2012: 352) observed: 
> ?Most studies use a Bayesian framework for estimating divergence dates 
> with probability curves between minimum and maximum bounds? but there 
> is presently no practical way to estimate curve parameters [i.e. 
> priors]?. One review of recent studies showed that the parameters are 
> usually not justified (Warnock et al., 2012). Another review noted 
> that ?current practice often consists of little more than educated guesswork? (Lee and Skinner, 2011:
> 540). Parham et al. (2012) wrote: ?the fact that a widely applied 
> methodology is subjected to such ambiguous assumptions that have a 
> major impact on results? is a major limitation of molecular divergence 
> dating studies? (p. 352).
>
>    Wilkinson et al. (2011: 28) also criticized the current approach, 
> as adopted in studies on primates (Chatterjee et al., 2009). They
>
> wrote the young age estimated for the main divergence in primates 
> (strepsirrhines vs. haplorhines):
>
>
>
> may be due to the use of two exponential distributions? These 
> distributions? implicitly assume that the true age is
>
> close to the minima [i.e. oldest fossil age] and unlikely to be much 
> older than those minima. This assumption, we feel, is unlikely to be 
> warranted, as it does not take account of the sizable gaps that exist 
> in the primate fossil record.
>
>
>
>
> On Fri, Jun 13, 2014 at 1:54 PM, Ken Kinman <kinman at hotmail.com> wrote:
>
> > John,        I found Jason's critiques to be quite insightful.  I would
> be
> > particularly interested in a straightforward response from you to 
> > his question about phylogeny calibration using priors (as follows):
> > ?Anyway, you contend that ?... the point here is that the 
> > misrepresentation of minimal ages as actual or maximal has led to
> erroneous
> > results.? I am sure that you are aware that phylogeny calibration 
> > allows for priors to be used and that this is common practice to use 
> > log priors
> to
> > accommodate not just for the uncertainty in the stratigraphy of 
> > fossils
> but
> > also for origination uncertainty. So is your complaint against 
> > someone in particular who used a very tight prior or about calibrations in general??
> > -------------------------------------------------------------------
> > _______________________________________________
> > Taxacom Mailing List
> > Taxacom at mailman.nhm.ku.edu
> > http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
> > The Taxacom Archive back to 1992 may be searched at:
> > http://taxacom.markmail.org
> >
> > Celebrating 27 years of Taxacom in 2014.
> >
>
>
>
> --
> Dunedin, New Zealand.
>
> My recent books:
>
> *Molecular panbiogeography of the tropics.* 2012. University of 
> California Press, Berkeley. 
> www.ucpress.edu/book.php?isbn=9780520271968
>
> *Biogeography of Australasia:  A molecular analysis*. 2014. Cambridge 
> University Press, Cambridge. www.cambridge.org/9781107041028 
> _______________________________________________
> Taxacom Mailing List
> Taxacom at mailman.nhm.ku.edu
> http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
> The Taxacom Archive back to 1992 may be searched at:
> http://taxacom.markmail.org
>
> Celebrating 27 years of Taxacom in 2014.
>
_______________________________________________
Taxacom Mailing List
Taxacom at mailman.nhm.ku.edu
http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
The Taxacom Archive back to 1992 may be searched at: http://taxacom.markmail.org

Celebrating 27 years of Taxacom in 2014.