[Taxacom] Methodological plurality [was cladistic analysis for morphological characters -- UPGMA is not cladistics]
Richard.Zander at mobot.org
Mon Nov 19 13:18:08 CST 2012
See replies below:
Richard H. Zander
Missouri Botanical Garden, PO Box 299, St. Louis, MO 63166-0299 USA
Web sites: http://www.mobot.org/plantscience/resbot/ and http://www.mobot.org/plantscience/bfna/bfnamenu.htm
Modern Evolutionary Systematics Web site: http://www.mobot.org/plantscience/resbot/21EvSy.htm
UPS and FedExpr - MBG, 4344 Shaw Blvd, St. Louis 63110 USA
From: taxacom-bounces at mailman.nhm.ku.edu [mailto:taxacom-bounces at mailman.nhm.ku.edu] On Behalf Of Kipling (Kip) Will
Sent: Monday, November 19, 2012 11:32 AM
Cc: taxacom at mailman.nhm.ku.edu
Subject: Re: [Taxacom] Methodological plurality [was cladistic analysis for morphological characters -- UPGMA is not cladistics]
That line of reasoning makes using multiple methods a sensitivity analysis. It would be the equivalent of using only one alternative set of parameters. There is quite an extensive literature on sensitivity analyses, but I can't recall ever seeing such a limited test of parameter space being considered sufficient.
REPLY: Yes. Didn't know that. I'll study up. Thanks! Remember, however, that I disagree with that which we are modeling. Macroevolutionary transformations are processes that theoretically happen, sister-group relationships are not.
In any case, conflicting resolutions can each be highly supported in different analyses and when they are, should we to take that to mean neither is highly supported?
REPLY: Yes, if they are both apparently good analyses using different data or data sets, then neither result is highly supported. Use Bayes' Formula for each result using the little support from the other result as an empiric prior.
When they agree it tells us little since we expect that or perhaps they are both wrong, when they disagree strongly it tells us nothing at all about the result (maybe something about the methods) and when they disagree and are each weakly supported it's just redundant. The weak support measures in each analysis already told us not to be too confident.
REPLY: No no. Suppose one result says there is support for (AB)C of 90% (and therefore of 10% for (AC)B if (BC)A is unreasonable. And if the support for (AC)B from a different result is 99%, when with 10% as a prior, the support for (AC)B, given the first analysis as a prior, is by Bayes' Formula is 92%.
If one analysis says (AB)C is supported at 0.80 posterior probability, and another analysis using different data says 0.80, too, then by Bayes' Formula with 0.80 as prior, we get 0.94 as posterior. Thus two or more analyses using different data can increase a result to acceptable levels. This seems to me to be correct use of total evidence.
Kipling W. Will
Associate Professor/Insect Systematist
Associate Director,Essig Museum of Entomology
More information about the Taxacom