[Taxacom] quote of the week

[Pierre Deleporte]

John Grehan a écrit :
>> From: Pierre Deleporte [mailto:pierre.deleporte at univ-rennes1.fr]
>>
>> - that a model is "mathematical" or verbal makes no difference, the
>> mathematical model is simply more precise,
>>     
> If something does not exist in nature then I don't see it being very
> precise - moth or no math. 
>   
your verbal model (assumptions) doesn't exist in nature, just in your 
brain, and maths are precisely used to specify what is in your brain 
(logics, reasoning)
> Alignment is only one problem. The other is whether the four alternative
> sequences are cladistically comparable to morphology. 
any set of descrete characters carried by taxa is cladistically analysable
> What sequences are really derived 
those absent in your outgroups of reference (sequence or base in an 
aligned sequence), exactly like for morphology (or do you just 'nose' 
them?)
this is the "special
> is something that I think is not really understood. For
> example, where is the unique sequence coding that does along with the
> unique ear structure? 
>   
this is not the same problem at all
you don't know what sequence corresponds to your structure, don't ask 
the molecularist to tell you what structure corresponds to her sequence
only genetics of the development can shed some light on this
> You took the bait. That was irony on my part. The fact is that most
> biologists seem to take the view that the orangutan theory is wrong
> because most biologists hold that view. Bad science I know. 
>   
bad argument I mean
>> - and as for "laws of great numbers" (which has strictly no particular
>> link with phenetics, by the way),
>>     
> Yes it does - in my opinion. 
>   
this is not a question of 'opinion' (very bad science and argument)

please check out some day what phenetics is all about (clustering on the 
basis of overall similarity), its conditions of validity for phylogeny 
inference (clocklike processes), and competing approaches (parsimony and 
likelihood optimisations of topology for a data matrix given some model 
of character evolution), and 'laws of great numbers" (expecting to pick 
out signal from noise if there is any law involved), and whether these 
questions are logically dependent or orthogonal...

> Only if they are apomorphies. There is a widespread argument by
> molecular biologists that they are more likely right because they have
> lots and lots of characters to look at. 
>   
you are not obliged to retain flawed arguments, numbers will do nothing 
if the data are incongruent
- but are they?
and parsimony molecular analysis makes use of synapomorphies only
(are you still ignoring this? didn't you check this elementary point 
since years of discussion?)
> They matter if they are apomorphies. 
>   
just like molecular apomorphies matter (but it's not written on the 
data, putative synapomorphy is coded in the matrix, surviving ones you 
know only after the analysis - the clique problem should pop up soon)
> As said before, I think comparing sequences to biological features is
> like comparing apples and oranges. One is cladistic, the other is not. 
>   
oh yes, I was expecting this one
and the other data set is "phenetic", I guess...
unfortunately, 'cladistic data' make no more sense than 'phenetic data',
no more today than when you stated this years ago,
nonsense is still nonsense, whatever the number of speakers,
and whatever the number of repetitions by the same speaker

there is nothing to do with such an 'argument', just stating again that 
such a sentence corresponds to no definition of 'cladistic' or 
'phenetic' in the scientific literature
(always applied to optimisation methods, not characters)

hence your statement is not understandable by the scientific community,
unless you re-define your notions of 'cladistic' and 'phenetic' an 
intelligible way
(I suggest: using proper names)

(the next steps should be a priori polarisation, and then cliques, and 
then silence, if I remember well) .

> Have done so, 
not on this list as long as I remember 
> only others don't accept them as flaws. 
in our circles, this can lead one to reconsider his argument carefully, 
and possibly take some advice about concepts and methods and vocabulary... 
> And why not the
> other way around? - no molecularist has ever demonstrated a flaw in the
> morphological analysis of primates. They just assume so. 
>   
OK but I never said that there was a flaw in your morphological 
analysis, I just fear that you are hastily 'resolving' the contradiction 
by rejecting molecular analyses as a whole on the basis of flawed 
arguments (assuming aligment problems without demonstrating them) or 
non-arguments (like 'phenetic characters')

the crux anyway would consist in your analysing the molecular data 'the 
right way'
- hence you'd have to renounce to the notion of 'molecular phenetic 
characters', or make explicit what you mean by 'molecular cladistic 
character' 
even if dressed in improper terms, at least people could try and 
understand what you mean by considering your (explicit!) procedures step 
by step

maybe you could begin with just a small sequence, for training (all 
available for free on the web)
looking forward for this molecular analysis...

Pierre



-- 
Pierre Deleporte
Université de Rennes 1
CNRS UMR 6552 
Station Biologique de Paimpont
F-35380 Paimpont   FRANCE
Téléphone : 02 99 61 81 63
Télécopie : 02 99 61 81 88