2. The Huybers normalization does the “one factor” change to remove off-centering ONLY.

3. It’s interesting to look at all the curves: mean, MBH98 (off-centered with variance norm), Huybers (on-center, with variance norm) and MM (on-center, no variance norm). They are all different.

4. I take the H comment and paper much more neutrally then Steve does. To me it shows different ones that you could pick from, both with less HS then MBH. But H is useful to show that “off-centeringness” only causes part of the difference between MBH and MM.

5. Your comment about the inappropriateness of using an iid-based statistic to normalize to get a corellation matrix sounds interesting and additive in the discussion. Is this a new point to science or have others discussed this danger in creation of correlation matrices? If so, who (from a theoretical stats perspecive with proofs and such)? If not, then you should put this into a fundamental stats contribution as this sort of subtle error could propogate into many other fields. ]]>

You want to prove the CO2 thing fine. But this is a specific discussion of PC techniques.

]]>It seems as if the page numbers were added after your reply:

“Thus, a third normalization is proposed where records are adjusted to zero-mean and unit variance over their full 1400 to 1980 duration, a standard practice in PCA [Preisendorfer, 1988 p22; Rencher, 2002 p393] here referred to as full normalization.”

]]>