Yes, but if your validation case says that tree ring size increases with temperature and you produce a correlation that says it is reducing this is a significant issue.

Is that what is going on here?

Re: Arthur Edelstein (#2)
“Multivariate regression methods are insensitive to the sign of
predictors.” — MBH
Cor(X, -Y) = -Cor(X, Y)
So if your predictor is upside-down, all that does is flip the sign of its estimated coefficient.

That makes sense, but at some point in building the proxy would not the sign matter?

I don’t know if you’ve got access through and organization or if you have the right cookies, but when I click on that link (and the M&M letter) I am told I need a subscription.

If you look at the table of contents, only some of the articles are listed as Open Access, and not the ones we’re talking about.

http://www.pnas.org/content/early/by/section

(link may not be static)

My understanding is that PNAS articles are freely available after some waiting period (six months?) but only selected ones are Open Access immediately.

This is strange. Now when I go to PNAS to view MM09, I get the page where they want to charge me $10.00 to read it. I do not understand why sometimes I can view it and sometimes I am asked to pay for it. How do they decided what is open-access and what is not? Do they use the Heisenberg uncertainty principle?

Sorry. I stole that joke from my son.

I know what open-access is but did not think PNAS was open-access. I now see I was wrong.

Is this open access?

http://www.pnas.org/content/early/2009/02/02/0812936106.full.pdf+html