[Annelida] re: eunicid phylogeny
kfitzhugh at nhm.org
Wed Mar 22 23:57:23 EST 2006
Torsten et al's reply is interesting, if for no other reason than they say nothing in defense of partitioned analysis. Resorting to claims that I am expressing opinions and not facts [sic] is no defense. I was not expressing an opinion. Rather, the basis of my claim is derived from the established tenets of non-deductive reasoning, from which there is no allowance for partitioned analyses. Torsten et al. do not provide a basis for why such analyses should then be acceptable. And there have been no legitimate arguments presented in the literature for partitioned analyses that acknowledge the accepted principles of non-deductive reasoning. That is not my opinion, but a consequence of what it means to be rational.
The claim for the superiority of likelihood [sic] is founded on serious misconceptions, not the least of which is misuse of the concept of likelihood and a misunderstanding of parsimony in abductive inference. To make any claims regarding likelihood against parsimony requires that one understand the actual inferential mechanics involving these concepts. Likelihood addresses the relation between evidence and hypothesis (support), parsimony addresses the relation between one's causal questions and hypothesis. Since likelihood can only be ascertained subsequent to an inference, it cannot be used as a criterion in the inference itself. Likelihood and parsimony deal with entirely different matters. So, my perspective is not derived from either parsimony or likelihood, contrary to the claim made by Torsten et al. Rather, my perspective is derived from the framework of the inferential process needed to produce explanatory hypotheses. Parsimony and likelihood are secondary issues.
Within the framework of Bayes theorem there is no allowance for the abductive inference of hypotheses. Bayesianism deals with changes in belief in hypotheses subsequent to confirmation by relevant evidence. This is an action wholly different from hypothesis inference. The former is inductive, the latter abductive. I discuss this in my Zootaxa paper and in the following paper: Fitzhugh, K. 2005. Les bases philosophiques de l'inférence phylogénétique: une vue d'ensemble. In: Deleporte, P. and Lecointre, G. (Eds.) Philosophie de la systématique. Biosystema 24: 83-105. I would be happy to forward these papers to anyone who wishes to see them.
None of the papers cited in Torsten et al's reply provide the relevant treatments of abduction to show the problems they think exist, much less provide a defense of partitioned analysis. Systematists continue to mistakenly think that Karl Popper (e.g., de Queiroz & Poe) or statistical methods (including bootstrap and Bayesianism) have anything at all to do with hypothesis inference or evaluation. Popper did not care how hypotheses are inferred, and statistics only addresses the testing of statistical (not explanatory) hypotheses, not their inference.
I do not care if Torsten et al. find my comments inflammatory. Invoking such an excuse avoids dealing with the real issues, much in the way they attempt to misrepresent the concepts of opinion and fact (NB: opinions ARE facts). A list serve is intended to promote open discussion. Let's argue the issues - that is the foundation for progress. The matter of partitioned versus total evidence is nothing new, so it is quite relevant to discuss the problem in this venue.
J. Kirk Fitzhugh, Ph.D.
Curator of Polychaetes
Invertebrate Zoology Section
Research & Collections Branch
Los Angeles County Museum of Natural History
900 Exposition Blvd
Los Angeles CA 90007
e-mail: kfitzhug at nhm.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Annelida