# Re: Dinosaurimony (Re: Definitions...)

```        And here I didn't want to start a big to-do over this. Let's wrap
this up fast, boys and girls, so we can get back to the big
scaley/furry/feathery guys.

George Olshevsky wrote (over two posts):
>Ah, yes. But nowhere does znc14@TTACS.TTU.EDU assert any kind of relativity
>for his "parsimony."
Actually, I didn't really even define parsimony. However, I hardly
think that there is any way to interpret the phrase "The simplest
explanantion which accounts for the most data with the least number of
unsubstantiated hypotheses and unsupportable assumptions is the most likely
to be correct." as anything other than relative. This statement requires a
*choice* to be made, and a choice implies the existance of multiple
hypotheses. The hypothesis which best meets the criteria of parsimony is the
prefered hypothesis.

>Unfortunately, accounting for the most data, having the least number of
>unsubstantiated hypotheses and unsupportable assumptions, and being the
>simplest explanation are mutually independent, so a parsimonius hypothesis may
>be impossible to formulate.
Independant? Not the way I formulated the statement quoted above.
The "simplest" hypothesis WHICH MEETS CERTAIN CRITERIA is the most likely to
be correct. therefore, we handle the criteria first. Therefore "simplest" is
dependant on the prior evaluation of the other three measures.
"Accounts for the most data" appears to be next most important, and
I do not think it is independant of unsubstantiated hypotheses and
unsupportable assumptions. An unsupportable assumption or an unsubstantiated
hypthesis is nothing more than an information claim with no evidence to back
it up. You (well, in this case, not you specifically, George) might very
well assert that the first is worse than the second, since it cannot be
backed up. Not accounting for data is almost the opposite, evidence with no
information claim. Evaluation of the weighting of these three criteria is
left up to the methodology which the worker or the discipline ascribes to
(e.g. in cladistics, each individual case is often weighted the same as an
attempt to keep the analysis "assumption free"). Once these have been
accounted for, overall simplicity may be assessed.
In any case, even if these criteria were independant, they could
still be used to evaluate the relative parsimony of differing hypotheses. As
for whether this makes a "parsimonious hypothesis" impossible, see Chris

>The point being that one must somehow measure simplicity versus data fitting
>versus unsupported assumptions in order to determine parsimony.
As noted above, no, you'd evaluate the "inner three" criteria first,
then simplicity.

>The use of parsimony in cladistics is simply one way to convert a character
>matrix into a cladogram. Nothing more or less. Here is a character matrix,
>here is the cladogram we get if we use parsimony analysis on the character
>matrix. If you want to interpret the cladogram as some kind of phylogeny, or
>want to use it to classify organisms, go right ahead--but beware.
This sounds like the "transformed" cladist view of things. My
problem with this sort of view is that, in order to NOT interpret a
cladogram as the best model for the phylogenetic relationships of organisms,
you must invoke assumptions about the evolution of the group (e.g. this
character is convergent, not homologous). In that case, your new phylogeny
has several more unsubstantiated hypotheses relative to the original
cladogram, and is therefore less parsimonious. Since parsimony is a logical
criterion for choosing between hypotheses, you should reject the perception
that alternate hypotheses are as likely to best approximate the true
That said, I have heard of some cladograms being rejected in favor
of cladograms with an equal (or slightly greater?) number of steps due to
subjective assessment of the likelihood of certain events. This is actually
not so bad, IMHO, since you are still using parsimony in your analysis, you
are simply suggesting that some aspect of the analysis appears to have been
weighted improperly. This is a testable hypothesis (get more data, run it