Let's be clear: this Tingley/Huybers paper is a proof-of-principle of an entirely new method of reconstructing past climate. It is not intended to be a "confirmation" of the MBH hockey stick, which Martin Tingley was very clear to tell me, though the preliminary results obtained so far have many similar features. Nor is it intended to solve every problem in the field of paleoclimatology, let alone purported problems in the data. (From my reporting, I think scientists believe that that data is pretty good, though data in any scientific field can _always_ be improved, and _is_ always being improved). In any case, this new result ought to, I think, damp criticism that the PCA approach was somehow unsound or flawed, as some have implied.
By the way, I asked Wegman for his thoughts on this new method, but he did not respond.
David
ReplyDeleteBut if the new paper incorporates Mann's PC1, then it essentially still uses Mannian short-centred PCA, right?
Bishop Hill wrote:
ReplyDelete> David
> But if the new paper
> incorporates Mann's PC1, then it
> essentially still uses Mannian
> short-centred PCA, right?
BH, the entire point of my article is that this new method *does not* use principal component analysis. It is independent of MBH's analysis.
That's exactly why is it important. It is a completely independent mathematical technique, completely new.
And it gives the same basis result.
it is like you have to speak slowly, to make yourself understood.
ReplyDeleteAs the bishop says, the new paper uses Mann's PC1 as one of its data series. The PC1 was obtaine by PCA. If one of the inputs for the analysis was obtained by PCA, then you ARE using PCA by definition.
If the PCA product then goes through another statistical transformation, it doesn't matter; you are still using principal component analysis.
In case you haven't noticed, there have been multiple criticisms of the methodology here. One is that the PCA methodology is sometimes dodgy, but another is the use of cherry-picked series that have a hockey-stick shape to begin with. If you select series that have a hockey-stick shape to start with, then you get a hockey stick out no matter what the analysis.
And one of the key issues is whether the series are cherry-picked. Why use Yamal, as opposed to its nearby neighbour which has not hockey-stick shape ?
I guess your intellectual position is that is is just some fantastic coincidence.
good luck
per
David Appell,
ReplyDeletePCA is NOT flawed. Although it was used questionably in Mann's early papers, the essential problem is that it is a data mining tool. It is intended to find possible signals, not test if they are real or significant.
The correctness (or incorrectness) of Mann's analysis can be authoritatively determined by seeing how well the proxies in PC1 correlate with temperature after both are updated.
It turns out that the original MBH hockey stick is a false signal. Let me be clear about what I mean by that:
PCA simply returns a series of possible signals. Mann identified one of these, which he claimed was a temperature signal.
Many of the data sources that Mann used have since been updated. Most importantly, the most heavily weighted proxy used by Mann (the gray bill bristlecones) were updated by Ababneh through 2003.
If we update the data with ONLY the Ababneh data and more recent temperature data, PCA no longer returns anything remotely similar to Mann's hockey stick.
So, irrespective of "true" historical temperatures, Mann's central thesis (that the proxies he identified are correlated with temperature as specified in PC1) was wrong.
I also note that if the goal of this paper is to discuss a new method of analysis, the authors demonstrated extremely poor judgement by using data series which have known issues.
Either the authors of the paper are stupid, or their motivations are not as you suggest in this post
More pseudo scientific garbage. A total misuse of math. There is no hockey stick.
ReplyDeleteLet us know, per, if your letter outlining the problems of this paper gets published. Surely your track record will be a good mark in the letter's favor.
ReplyDeleteToodle-pip!
Best,
D
Dano
ReplyDeletewhat a pleasure to read your non-sequiturs again !
so tell me, what is a job in green infrastructure ? How much financial interest do you have in these discussions, and in your eco-evangelism ?
just wondering.
per
"96 MXD series out of a total of 116 proxy series in the Tingley-Huybers network go down, but the overall reconstruction goes up. Hmmmm."
ReplyDeletei think it is a huge advance to be credited to mcintyre, that hockey-stick productions are now so well understood and so easy to refute. that good journals the chance to debunk them before publication and embarrasing their own integerity.
per,
ReplyDeleteWe note that you still must change the subject away from the fact that your argument is content-free, this time to the tired green conspiracy.
One hopes you continue your pattern when you write your letter outlining the problems of this paper - that will ensure it gets published, certainly. Surely your track record will be a good mark in the letter's favor.
Best,
D