scientific omissions and published negative evidence

[James Francis Lyons-Weiler]

Harvey Ballard brings up an interesting point, one that is not restricted
to molecular systematics, namely the publishing of negative results.

The same problem exists in ecology - experiments have been performed a
couple of times, and finally some one at a meeting will say something like
"oh, we tried that - and it didn't work".  At the core of this
problem is the question of statistical power; i.e., did the
researcher have a large enough sample size to actually reject the
null hypothesis if it were in fact false?  If so, then _maybe_ they
should assert that the experimental treatment had no effect.

There is a difference, however, in using nrDNA ITS sequences and
finding no variability in one group and then expected those results to
have any bearing on the utility of ITS or its informativeness in other
groups.  It is a fallacy to expect that just because it is informative in
one study that it will be informative at another (this is true for any
region, mtDNA, genomic or cpDNA) because rates vary among lineages and
taxonomic groups by rank are not commensurate (a species in Astragalus is
not the same as a species in Lomatium).  Therefore, authors who
generalize about the utility of a regions are making risky
generalizations, and those that expect their organisms to carry the same
level of signal at that locus are taking a chance that it might be
informative.  The primary reason why people try out regions that others
have found informative is that they can get primers for them, and that
the perceived probability of informativeness is taken as real.

Although there is this difference, the question of the power of the test
obtains here, too.  Researchers can do themselves a favor by asking how
many more nucleotides or amino acids appear to be needed to maximize the
signal:noise ratio if they could sample more with the same level of
information. I have an ms in the works describing this application; others
have tree-based ways of approaching the same question, but they conflate
BP proportions with signal.  The option is live in rasa 2.1 (power &
effect) and is described in the manual.  At this point I'm looking for
empirical data sets to showcase in the ms, so anyone interested in having
me address this for their data (old or new) should contact me off the
list.

James LW
_______________________________________________________________________________

  \  /   /    \  /           JAMES LYONS-WEILER           ______________
   \/   /      \/                                        |..............|
    \  /       /                                         |..............|
     \/       /              DOCTORAL PROGRAM IN         |..............|
      \      /               ECOLOGY, EVOLUTION, AND     |...***........|
       \    /                CONSERVATION BIOLOGY        |..*****.......|
        \  /                                             |.******.......|
         \/                  1000 VALLEY ROAD/186        |********......|
    ______________           THE UNIVERSITY OF            --------------
   | will perform |          NEVADA, RENO
   |  statistical |          RENO, NEVADA 89512-0013
   | phylogenetic |
   | analyses for |         "(Biology) is not religion; if it were, we'd
   |    food      |          have a much easier time raising money."
    --------------                       -Leon Lederman


                EXPLORATORY ANALYSIS OF PHYLOGENETIC DATA
                RASA 2.1 SOFTWARE FOR THE MAC
                Download @
                http://loco.biology.unr.edu/archives/rasa/rasa.html
_______________________________________________________________________________