Search the Taxacom archive · Browse by month · Mirror of the former lists.ku.edu archive (1992–2025)

[Taxacom] Molecular phylogenetic hypotheses

Kirk Fitzhugh kfitzhug at nhm.org
Wed Aug 18 01:21:56 CDT 2010 

Fortunately, we are not destined to such nihilism: http://spot.colorado.edu/~cleland/articles/Cleland.PS.Pdf, http://spot.colorado.edu/~cleland/articles/Cleland.Geology.pdf (to name just a couple of examples). There's nothing mysterious about hypothesis testing. Other than how the topic was contorted in the cladistics literature from the '70's to the present.



-----Original Message-----
From: Don.Colless at csiro.au [mailto:Don.Colless at csiro.au]
Sent: Tue 8/17/2010 10:32 PM
To: Kirk Fitzhugh
Cc: taxacom at mailman.nhm.ku.edu
Subject: RE: [Taxacom] Molecular phylogenetic hypotheses
 
Several folk have mentioned "testing" of phylogenetic hypotheses; but in what would such a test consist?  Back in the '30's R. J. Tillyard had a solution: Cretaceous dragonflies used to flutter around his bedroom every night, allowing immediate access to vital morphological data. We were tolerant of eccentrics in those days; but no one believed him! Short of such access to history, phylogenetic hypotheses have the problem that they have no logical consequences. They must remain an "inference to the best conclusion". 
 
Donald H. Colless
CSIRO Div of Ecosystem Sciences
GPO Box 1700
Canberra 2601
don.colless at csiro.au
tuz li munz est miens envirun
________________________________________
From: taxacom-bounces at mailman.nhm.ku.edu [taxacom-bounces at mailman.nhm.ku.edu] On Behalf Of J. Kirk Fitzhugh [kfitzhugh at nhm.org]
Sent: 18 August 2010 11:21
To: Stephen Thorpe
Cc: TAXACOM
Subject: Re: [Taxacom] Molecular phylogenetic hypotheses

  It's not a matter of 'more complete' versus something lesser. It's a
matter of the relations between the basis for inferring some hypothesis
given some set of evidence. In the event there is relevant evidence
that's being actively partitioned out of an inference, then there is
less of a rational basis for accepting that conclusion/hypothesis. Worse
yet, actively comparing branching diagrams is a meaningless exercise -
the two diagrams don't matter to one another, since each only applies to
the premises/evidence used in the inference.

Re "Think of it this way: a subset of data (say morphological) points to
a conclusion which you don't accept if it contradicts that of another
subset (say molecular), but does adding more data necessarily help?"
What's the empirical basis for accepting the conclusion beyond the
premises used to make the inference? The conclusion can't 'contradict' a
conclusion derived from some other class of observations. The two
conclusions are irrelevant to one another. Don't confuse a simple
branching diagram with an explanatory hypothesis.

"Why is that conclusion any more rational to believe than any
contradictory subset?" Rationality is a relation between premises and
conclusion. Conclusion A is more rational than B if it is the case that
A was inferred from one's total available relevant evidence and B was
only inferred from a subset of that evidence.

"The BIG assumption here seems to be that subsets point to the wrong
conclusions simply because they are only subsets (i.e., incomplete), and
any conclusion that the complete data points to is to be preferred on
the grounds that it is complete data. But perhaps both incomplete and
complete datasets can point to the wrong conclusions for any number of
reasons unconnected to completeness??"

Just like 'truth' throwing around 'right' and 'wrong' is just not
appropriate to this issue. We're only talking about going from observed
effects to a hypothesis that offers an impossibly vague explanation in
the form of a cladogram. To speak of whether or not some hypothesis is
'right' or 'wrong' only matters at the point one goes out and actually
follows the established mechanics of putting that hypothesis to the test.

On 8/17/2010 6:00 PM, Stephen Thorpe wrote:
> yes, I was simplifying it a bit for brevity, but substitute "rational
> basis for belief" for "truth" and the question still stands:
> does a more complete set of data give you any more rational basis for
> belief in a conclusion? Think of it this way: a subset of data (say
> morphological) points to a conclusion which you don't accept if it
> contradicts that of another subset (say molecular), but does adding
> more data necessarily help? Suppose we have "complete data" and it
> points to a particular conclusion. Why is that conclusion any more
> rational to believe than any contradictory subset? The BIG assumption
> here seems to be that subsets point to the wrong conclusions
> simply because they are only subsets (i.e., incomplete), and any
> conclusion that the complete data points to is to be preferred on the
> grounds that it is complete data. But perhaps both incomplete and
> complete datasets can point to the wrong conclusions for any number of
> reasons unconnected to completeness??
> ------------------------------------------------------------------------
> *From:* J. Kirk Fitzhugh <kfitzhugh at nhm.org>
> *To:* Stephen Thorpe <stephen_thorpe at yahoo.co.nz>
> *Cc:* Bob Mesibov <mesibov at southcom.com.au>; TAXACOM
> <taxacom at mailman.nhm.ku.edu>
> *Sent:* Wed, 18 August, 2010 12:24:42 PM
> *Subject:* Re: [Taxacom] Molecular phylogenetic hypotheses
>
> There are two answers to your query: (1) there is no guarantee any
> inference will give one a true [sic] conclusion (except in deduction);
> (2) the notion of truth is irrelevant to the issue (the matter at hand
> is one of having a rational basis for accepting a particular
> conclusion based on available premises/evidence). Talking about
> 'truth' only comes into the mix when one actually tests hypotheses -
> something that is almost never done with phylogenetic hypotheses.
>
> On 8/17/2010 4:45 PM, Stephen Thorpe wrote:
>> this raises something I have been wondering: in general, if different
>> subsets of data point to different results, why would the "total
>> evidence approach" (i.e. the complete set of data, or at least a more
>> complete subset of data) be expected to necessarily point to the
>> "truth"???
>>
>> ------------------------------------------------------------------------
>> *From:* Bob Mesibov <mesibov at southcom.com.au>
>> *To:* kfitzhug at nhm.org
>> *Cc:* TAXACOM <taxacom at mailman.nhm.ku.edu>
>> *Sent:* Wed, 18 August, 2010 11:37:26 AM
>> *Subject:* [Taxacom] Molecular phylogenetic hypotheses
>>
>> Hi, Kirk.
>>
>> Many thanks for the draft. I'm copying this portion of my reply for
>> Taxacom because there's what might be seen as a counter-example in
>> the recent literature, which Taxacomers might like to read.
>>
>> You make the point that a typical molecular systematics research
>> stream consists of building a phylogeny from sequence data,
>> publishing it and moving on (to either another group or the same
>> group with more data) to do the same thing again and again, with no
>> attempt to understand the result. This freely available paper
>>
>> http://rspb.royalsocietypublishing.org/content/early/2010/08/05/rspb.2010.0590.full.pdf+html
>>
>> does something more. The authors examine the possibility that the
>> disagreements between certain published phylogenies are due in part
>> to *analytical biases*. They apply very simple tests which support
>> this idea.
>>
>> What is surprising and disappointing to me is that one or more of
>> those tests might have been done earlier by any of the research
>> groups working on arthropod phylogeny. They weren't. For some
>> molecular systematists, the way to get better phylogenies is to use
>> more or longer sequences and more terminals. The hope here is that if
>> you throw more data at the problem, any systematic biases will be
>> swamped by 'signal'. If there are disagreements between research
>> labs, you *suspect* bias but you don't test for it, maybe because
>> your previously published work would be revealed to be... less than
>> useful.
>>
>> Perhaps for some molecular systematists, the analytical method of
>> choice is an oracle whose pronouncements cannot be questioned, only
>> published. No wonder you see that sterile sequence-analyse-publish
>> repeat. The same oracle-worship is seen in *choice* of inference
>> method. In a post earlier this week, Thomas Simonsen compared
>> neighbour-joining 'phylophenetics' unfavourably with
>> other-inferential 'phylogenetics'. Thomas might be right in the
>> particular case he's considering, but distance-based trees are often
>> very close to maximum likelihood trees for closely related taxa (you
>> might say 'close within the uncertainty of the method'). Should you
>> not do distance-based analyses because of a philosophical distaste
>> for them, even though they give the same result as another method?
>>
>> [Hoping this doesn't spawn a long thread, off to the field for
>> several days shortly]
>>
>> Best wishes,
>> Bob
>> --
>> Dr Robert Mesibov
>> Honorary Research Associate
>> Queen Victoria Museum and Art Gallery, and
>> School of Zoology, University of Tasmania
>> Home contact: PO Box 101, Penguin, Tasmania, Australia 7316
>> 03 64371195; 61 3 64371195
>> Webpage: http://www.qvmag.tas.gov.au/?articleID=570
>>
>> _______________________________________________
>>
>> Taxacom Mailing List
>> Taxacom at mailman.nhm.ku.edu <mailto:Taxacom at mailman.nhm.ku.edu>
>> http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
>>
>> The Taxacom archive going back to 1992 may be searched with either of
>> these methods:
>>
>> (1) http://taxacom.markmail.org <http://taxacom.markmail.org/>
>>
>> Or (2) a Google search specified as:
>> site:mailman.nhm.ku.edu/pipermail/taxacom  your search terms here

_______________________________________________

Taxacom Mailing List
Taxacom at mailman.nhm.ku.edu
http://mailman.nhm.ku.edu/mailman/listinfo/taxacom

The Taxacom archive going back to 1992 may be searched with either of these methods:

(1) http://taxacom.markmail.org

Or (2) a Google search specified as:  site:mailman.nhm.ku.edu/pipermail/taxacom  your search terms here

More information about the Taxacom mailing list