date: Wed, 05 Nov 2003 07:53:15 -0500 from: "Michael E. Mann" subject: ad additional suggestion... to: Scott Rutherford , "Raymond s.bradley" , mhughes@ltrr.arizona.edu, Tim Osborn , "Keith Briffa" , Phil Jones , mann@virginia.edu Dear All, Hopefully, the JGR paper on which Scott is first author and we're all co-authors, should be coming back from review soon (Scott--please contact J. Climate to find out what the status is ASAP). As I mentioned before, I see this as a natural first step in the broader future collaborative effort that Keith has nicely layed out, in which we can look in detail at the sensitivity to selection of candidate predictors, issues of seasonal, spatial sampling, etc---all of the things we all know need to be looked at in more detail. I strongly endorse the idea of making this a collaborative effort of the full group of us, perhaps w/ Tim and Scott in the lead of the joint project (do people feel this is reasonable?). Between the two groups, I think we're fully funded for this type of activity... Along these lines, I have a suggestion for Scott regarding the J. Climate paper that should be coming in from review soon. A few important measures taken here can go a long way to combatting the latest E&E criticism of MBH98, since we get essentially the same results for the MBH98 network w/ a completely different statistical method, and explicitly compare results w/ other networks, etc. By the time the paper appears, we want to have a supplementary website (to which we should plan to refer in the paper!) that will have *all* data, and *all* codes (Scott's MATLAB codes--clean these up first though Scott) used, and a *thorough* description of all methodological details (no matter how small) so that independent scientists would have everything they need to reproduce the results. We're not all in the habit of doing this, and its now clear that, in certain cases, we need to... I also have one other suggestion--Scott, you should go through the MBH98 dataset (refer to the original description to determine where the termination dates were) and make sure that any extensions beyond the last available data point by persistence that I performed originally are removed--we don't need them, since RegEM can handle the missing data in estimating the required covariances anyway! You should also do an experiment where the MBH98 network is only used in calibration through 1971 (the earliest date for which no series have been extended to the 1980 boundary by persistence), since the "PC" predictor series are already based on some data that have been extended, and its not worth the bother to redo these all. Stopping the calibration in 1971 is another way of avoiding the use of any persistence-extended data... In our reply to MM03, I'll be showing that we get a virtually identical result if we only use a 1902-1971, rather than 1902-1980, training period, taking away from MM03 the argument that the extension of some series by persistence to 1980 makes any difference. If we do all of the above for the in-review Rutherford et al J. Climate paper, and have the website up and running (perhaps working w/ Mark Eakin at NGDC to have the webpage located there to, or at least a link to our webpage), we take away a major source of criticism. Thoughts? mike ______________________________________________________________ Professor Michael E. Mann Department of Environmental Sciences, Clark Hall University of Virginia Charlottesville, VA 22903 _______________________________________________________________________ e-mail: mann@virginia.edu Phone: (434) 924-7770 FAX: (434) 982-2137 [1]http://www.evsc.virginia.edu/faculty/people/mann.shtml