Re: stats

[Stan Friesen <swf@ElSegundoCA.NCR.COM>]

From: "Bonnie Blackwell, x 3332" <bonn@qcvaxa.acc.qc.edu>

 > Stan's posting (9612.03) oversimplified the problem of picking 75%
 > f the lancian species.  No one could imagine that the selection was
 > truly random.

True, not entirely random.  But I am not yet convinced that the data
can be *distinguished* from an essentially random process. (Which
is a slightly different statement).

 >  And each group would have experienced different
 > selection (extinction) pressures as a whole, as would each family
 > and order and so forth.  The problem is in picking 75% overall, but
 > maybe only 30% mammals, 100% dinos and pteros, 60% birds, etc....

This is exactly the result one might get from a random process.
That is my main point - the particular distribution of extinctions
may not be any different than what would be produced by a random
process.

The issue here is one of using the data to support a model.  If the
data is indecisive, the random model cannot be rejected.

My suspicion is that there was some non-random component - some bias -
in the extinctions, but that there was also a substantial stochastic
component.  This is what I really mean when I say that the dinosaurs
were not *specifically* singled out by the extinction mechanism. One
of the other groups with, say, >= 50% extinction might equally well
have died out.

 > It may be (or more likely is not) rather random within narrowly
 > defined groups, but certainly not across the board.

Actually, I think it is *very* likely that it was partially random
within fairly ecological groups.  How to *test* this model against
the data, however, is a difficult issue.

Now, if a proper statistical analysis of the model I presented compared
against the actual data shows a stistically significant variance from
the expected pattern, then the *broad* stochastic model can be rejected.
Of course this still leaves various combination models.

swf@elsegundoca.ncr.com         sarima@ix.netcom.com

The peace of God be with you.