[asa] Re: The Hockey Stick Holds Up

From: Rich Blinne <rich.blinne@gmail.com>
Date: Wed Sep 10 2008 - 15:38:34 EDT

 On Wed, Sep 10, 2008 at 12:06 PM, j burg <hossradbourne@gmail.com> wrote:
> On page 293 ff of PHYSICS FOR FUTURE PRESIDENTS, Muller argues that
> the "hockey stick" graph is fundementally flawed. Does this latest
> refute his arguments?

Yes. Muller's criticism was the use of principle component analysis (PCA) in
Mann 98. Muller puts it this way:

> math questions involve the procedures for combining data sets. Mann used a well-known approach called principle component analysis. This method extracts from a set of proxy records the behavior that they have in common. It can be more sensitive than simply averaging data, since it typically suppresses nonglobal variations that appear in only a few records. But to use it, the proxy records must be sampled at the same times and have the same length. The data available to Mann and his colleagues weren't, so they had to be averaged, interpolated, and extrapolated. That required subjective judgments which—unfortunately—could have biased the conclusions.

The current study doesn't use PCA. Dr. Gavin Schmitt explains the current
methodology as follows:

There is no separate PCA stage in this analysis. Both the CPS and EIV
methodologies have their own ways to deal with statistical redundancy
(i.e. making sure that several nearby and similar records don't get
overweighted in the final reconstruction). For CPS, it is done through
gridding onto a 5×5 grid prior to the CPS procedure, while EIV takes
account of the co-variance of individual proxies directly. This is
better explained in the SI (linked above).

CPS is composite from scale and EIV is error in variables. The supporting
information that describes the techniques is here:


One other thing to note is the other that is controversial about Mann 98 is
the use of tree ring proxies. This study found the same answer even without
using the tree proxies. Mann et al take good advantage that the reliability
and number of proxies have increased since 1998. So, they can throw away the
tree ring proxies and still get good information. This also brings up
another point. The skeptics often attack Mann 98 only and the techniques
used which have been refined over the last 10(!) years. Take a look at the
graph again. Note the other studies cited in it. Many of these use different
techniques to compute the results yet come up with the same answer.


Rich Blinne
Member ASA

To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Wed Sep 10 15:39:26 2008

This archive was generated by hypermail 2.1.8 : Wed Sep 10 2008 - 15:39:26 EDT