Wednesday, December 21, 2011

1153172761.txt

From: "Cooke, Barry" <bcooke@NRCAN.GC.CA>
To: ITRDBFOR@LISTSERV.ARIZONA.EDU
Subject: Re: [ITRDBFOR] Joe Barton's hockey stick hearing coming up
Date: Mon, 17 Jul 2006 17:46:01 -0400
Reply-to: ITRDB Dendrochronology Forum <ITRDBFOR@LISTSERV.ARIZONA.EDU>

"Non-independence" of reconstructions and "worthlessness" of the hockey
stick model were raised as separate issues.

If the worth of a model is measured by its ability to predict, then a
model that explains 0.5% of the variation in some variable is fairly
(but not necessarily completely) "worthless". Surely, one hopes for
better. Especially where consensus is required.

The proxy data on which multi-proxy reconstructions are based may be
statistically independent, but the reconstructions themselves are not.
This is not because of any lack of "independence" (i.e. objectivity)
among networked researchers, but a measurable fact of arithemtic. To the
extent that multi-proxy reconstructions are built on the same proxy
data, they are statistically non-independent (i.e. correlated).

i.e. It's not the non-independence that make the model worthless. It's
the uncertainty.

On your last point of social networks, try a Google search of 'Exxon
Secrets'. The difference between a ruling orthodoxy and a scientific
network is not the degree of connectivity, but the mode of governance:
coercion & inculcation vs. facts & reason (including statistical
inference). Be wary of any science that loathes statistics or resents
external investigation. That's the start of rot.

If Wegman et al. are suggesting that statisticians should be put to work
to serve the interests of paleoclimatologists (which they are), then who
on this list is going to argue that? I say let's put them to work!

Barry Cooke

-----Original Message-----
From: ITRDB Dendrochronology Forum
[mailto:ITRDBFOR@LISTSERV.ARIZONA.EDU]
Sent: Monday, July 17, 2006 6:43 AM
To: ITRDBFOR@LISTSERV.ARIZONA.EDU
Subject: Re: Joe Barton's hockey stick hearing coming up

>Maryanne's message further claims that the "characterization of the
>hockey stick as 'worthless' underscores what appears to be a basic lack

>of understanding of how scientific consensus is formed". Yet if a
>consensus is based on invalid statistical analysis, then the consensus
>is wrong.

To explain my point (and my apologies to those to whom this is obvious):
it would not be unprecedented for a scientific consensus to be wrong.
However, there is also ample precedent for papers containing flaws
(which virtually all do, if somebody looks hard enough, or has the
misfortune of having the resources of Congress devoted to finding them)
to have constructive influence on debate. To take an example from
history, many of Charles Darwin's observations are pure amateurish
nonsense by the standards of even the late 19th century, but no one
would doubt their value in building the consensus for evolution. The
question is not always strict veracity, but whether work provokes
fruitful questions, or leads research in a constructive direction. (By
the way, this is not to take a position on the Wegman judgement on the
MBH papers).

>Dave's message further claims that there are multiple "independent
>lines of evidence" for the hockey stick. The Wegman report discusses
>this claim. See especially p.46-47, which cite twelve different
>studies and concludes that those studies "cannot really claim to be
>independent".

This part of the report is more precious than useful. In most empirical
fields, leading primary investigators have linkages--nothing unusual
about that. We could construct similar matrices of social networks in
physics, biology, statistics. That doesn't mean the works produced in
physics, biology or statistical theory are "worthless". A similar point
can be made about different investigators using the same proxy data. In
fact, isn't it one of the recommendations of the Wegman report that the
paleoclimate community share data more effectively? Seems that if that
recommendation was followed, certain statisticians would have even more
occasion to complain of a lack of true independence. Seems these poor
climate experts can't win!

Wouldn't it be interesting to see a "social network" matrix--or a
funding matrix--between those the scientists, statisticians,
Congressional Republicans, and oil companies most passionate about
"debunking" global climate change?


Dr. Maryanne W. Newton
Research Associate
Malcolm and Carolyn Wiener Laboratory for Aegean and Near Eastern
Dendrochronology Cornell University

No comments:

Post a Comment