Mail*Link(r) SMTP FWD>certainty/un

Meredith Lane writes:

-------------------------------------- Date: 28/07/1995 17:41 From: Meredith A. Lane

Mediocre spirits demand of science the kind of certainty which it cannot give, a sort of religious satisfaction. Only the real, rare, true scientific minds can endure doubt, which is attached to all our knowledge. -- Sigmund Freud

As an adolescent I ... craved factual certainty ... so I became a scientist. This is like becoming an archbishop so you can meet girls.

-- M. Cartmill

------------------------------------------------------------------------- And Warren responds:

Whether I am judged to be a "mediocre spirit" or quite similar to an "adolescent" [as might be concluded from the above quotes posted in response to my original posting quoted in the first few lines] is for everyone else to decide. It is very difficult to answer previous posts, and then have someone else nit-pick away at the details of my responses. I did not mean that I would not tolerate any uncertainty as a scientist; what I have been saying all along is that the total uncertainty in phylogeny reconstruction methods is so great that it exceeds my capacity to accept it. I might mention that I was originally trained as a mathematician, later as a statistician, and finally as a botanist, and I use statistical methods almost daily and certainly weekly in analyzing RAPD data (see my papers in PCR Methods and Applications 1994, Vol. 4: 31-37 and 4:38-43) and data from SSRs (simple sequence repeats). It is not uncertainty in phylogeny reconstruction methods that I object to. It is the inability to test the methods empirically against real data. I mean, even astrology takes real data (birth place, date, and time), carries them thru a complex and sophisticated procedure [beyond the ken of most of us, I am sure, just like phylogeny reconstruction appears to many], and comes up with predictions and recommendations. And astrology has the added benefit that it can be compared to reality to see how well it works! When I am wearing my statistician hat, I am generally concerned with minimizing two things: variance + bias. As I decease the variance, my estimates become more precise (have smaller standard errors) -- although the estimates themselves may differ wildly from the true value, if the estimates are biased. I would conjecture that as we pile up more and more data sets and different types of data from a specific phylogenetic group of interest we do decrease the variance, but we do not know how biased the results are, nor what we can do about it. Extinction has a tremendous effect on how well we are able to reconstruct phylogenetic trees, and I'll bet it has occurred in almost all but the very most recent plant groups. This source of difficulty can never be removed from the picture.

-Warren