[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: How to Train Your New Papers



> >  We attached a prior probability to each
> fossil-based minimum age,
> 
> Are there any maximum ages? (I don't have access to the
> paper here.) If not, it had to be _expected_ that many of
> the molecular dates would be too old. Minimum ages can only
> push molecular dates down (stratigraphically down), and when
> there's nothing to push them up, down is where they'll stay.
> Appendix 10 of my 2007 paper with Michel Laurin demonstrates
> this at excruciating length...

I'm still waiting for somebody who has done and seen it all to publish a 
gold-standard review on what can go *wrong* with mol-phyl studies and how to 
avoid it.

For birds, the name of the game these days seems to be multigene or even 
whole-mitogenomic studies. But given that gene trees do not equal species 
trees, this introduces an increasing probability of this particular error 
(which always leads to an overestimation of age). Essentially, without 
coalescent analysis, such studies are not very useful. Also concatenation is 
very shaky; it can save a lot of work, but before using it it must be checked 
whether it *can* be used in this case (it's at least very risky to concatenate 
indel-heavy sequences). Interesingly, the literature about things that can go 
wrong with non-molecular cladistic analyses is quite extensive.

Another thing that is overdue is incorporation of geography into cladistic 
analyses. Hard to quantify, and hasn't been much used for this reason. At first 
I thought about quantifying it via polar coordinates, but that's not good. 
Perhaps one could simply treat the general regions of an OTU's occurrence as 
characters, and analyse it via a network of high/medium/low/"virtually 
impossible" probability of dispersal to other regions (i.e. character 
transition). E.g. something that is found in Jurassic Europe has a high 
ppprobability of having relatives elsewhere in Europe, a medium probabilty of 
relatives in Asia, a low probability for Africa, and a close relative 
unexpectedly popping up in Australia would indicate either a significant ghost 
record or that you should recheck your analysis.

Plate tectonics would need to be accounted for, but that can be done. The 
interesting aspect is that it could be dealt with in a similar way as a 
substitution model in mol-phyl 
(http://en.wikipedia.org/wiki/Substitution_model), and would be open to ML 
bootstrapping (regarding ML bootstrap vs MP PP, see also 
http://mbe.oxfordjournals.org/cgi/reprint/20/2/248.pdf).


Regards,

Eike

__________________________________________________
Do You Yahoo!?
Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz gegen 
Massenmails. 
http://mail.yahoo.com