I haven't read PAGES 2017 before, but I think the issue is that all the non-tree-ring proxies have been "screened out" from the North American network prior to data analysis. This is detailed (with helpful maps) in the Supplementary Information: https://media.nature.com/original/nature-assets/sdata/2017/s.... As they apply successive layers of screening for factors such as "False Discovery Rate" and correlation with "Mean Annual Temperature", everything except the tree ring proxies drop out from North America (other than a single coral series in the Yucatan).
This is essentially the basis of McIntyre's criticism of the paper. Traditional statistical technique says that you should come up with criteria for a valid proxy in advance, screen without looking at the data, and then use all the data. Once you start screening out sites after peeking at the data that is going to be used for your analysis, it's hard to know if you've biased your results towards a preordained conclusion.
You should pause at this point and ask yourself: Is this really what the paper does? How do they justify that? What do outside statisticians think of this approach? Why isn't this talked about more prominently outside of the "skeptic" blogs?
If tree rings are to be ignored, wouldn't the right question be, what does the analysis look like using only the non-tree ring data for North America?
Yes, this is a fine question, although McIntyre isn't suggesting that all tree series be ignored, only a few problematic "stripbark bristlecone pines". In prior work, the answer is that if you leave out these few trees, the emblematic "hockey stick" graph disappears. It's possible to recover the sharp upward rise without the stripbarks, but only by adding in an "upside down" Finnish lake sediment proxy (upside down in the sense that the principle components analysis inverts the usual correlation of thick and thin layers to hot and cold).
This is what section of the blogpost with the subtitle "Background: Stripbark Bristlecones and Mann’s CENSORED Directory" is about. The "spaghetti graphs" are showing the contributions from the other groups of proxies. Reading the historical links in that section will give you a tour of McIntyre's argument. His argument is not that global warming is not happening, just that much of the proof being offered is statistically very flawed.
(It's worth noting the parallels between the criticism of post-hoc screening above, and the suggestion that the bristlecone pines should be screened out. Why one, and not the other? I don't think there's a clear answer.)
Could that imply that you have a bias error in your understanding of Eschenbach and McIntyre because your understanding is based more on the interpretations presented at ClimateAudit than from the primary literature?
As they apply successive layers of screening for factors such as "False Discovery Rate" and correlation with "Mean Annual Temperature", everything except the tree ring proxies drop out from North America (other than a single coral series in the Yucatan).
And yet the 'hockey stick' remains in all the analyzes. That is, figure 7 shows "Global composites for various binning intervals and screening criteria, as indicated in subplot titles". More specifically, "Screening options comprise: no screening (none), regional temperature screening (regional), or regional screening adjusted for the false discovery rate (regionalFDR)."
Once you start screening out sites after peeking at the data that is going to be used for your analysis, it's hard to know if you've biased your results towards a preordained conclusion.
The 'no screening' analysis is figure 7a, is it not? Which has a hockey stick.
Certainly. I actually started by reading MBH98 (after a friend interviewed Hughes) and I found it statistically very weak. I then discovered McIntyre's pre-ClimateAudit site at the time he was working on his rebuttal. Discovering someone else who was bothered by the same flaws (and was able to discuss them in a more rigorous form than I could) predisposed me to view him in a positive light. I'd like to think my continued positive impression of him has to do with the strength of his arguments since then, but probably at least some part is path dependent bias.
And yet the 'hockey stick' remains in all the analyzes.
The issues with post-hoc screening are separate from his initial criticisms, and are about statistical principles rather than the resulting graphs. McIntyre isn't saying that you only get a hockey stick if you screen out certain proxies. Rather, he thinks the analysis is overly sensitive to a small number of problematic "stripbark" series in the American West. If you include these series (or a small number of alternatives), you get a blade; if you leave them out, you don't. He believes that since most of the other proxies are only weakly tied to temperature, they average out to a mostly flat red noise that forms the historically flat "handle".
I don't feel qualified to make his full argument, though. While I tried to keep up with the primary literature at one point, I haven't been reading much lately, and I've never been as deeply into it as McIntyre himself. My goal was simply to say that I think that both he and Eschenbach have some good points to make, that both are interested in finding the truth rather than spreading misinformation, and that they should be read and considered rather than being rejected out of hand.
MBH98 was 20 years ago. https://en.wikipedia.org/wiki/Hockey_stick_graph goes into details about the various analyzes of MBH98 and later, including criticisms and counter-criticisms. The Moberg 2005 analysis I referenced above is one of them, which used wavelet transforms instead of PCA. It found similar conclusions as MBH99, the extension of MBH98.
> A 2010 opinion piece by David Frank, Jan Esper, Eduardo Zorita and Rob Wilson (Frank et al. 2010) noted that by then over two dozen large-scale climate reconstructions had been published, showing a broad consensus that there had been exceptional 20th century warming after earlier climatic phases, notably the Medieval Warm Period and Little Ice Age
It then adds that more reconstructions have been done since then, again following the broad consensus.
they should be read and considered rather than being rejected out of hand.
This goes back to my original objection, which is that the g'g' etc. poster didn't point to something to read, other than "Willis Eschenbach's work over on https://wattsupwiththat.com/ ".
I can certainly reject out of hand the undirected suggestion to read hundreds if not thousands of blog posts without having some solid reason to believe it's worthwhile, especially given that several of the poster's other comments on the topic seemed to be wrong, and internally inconsistent.
McIntyre's blog post starts "The PAGES (2017) North American network consists entirely of tree rings.".
The 2017 article "A global multiproxy database for temperature reconstructions of the Common Era" by the PAGES2k Consortium at https://www.nature.com/articles/sdata201788 says:
> While the M08 database contains 75% more records than this collection, these records are overwhelmingly land-based, from the Northern Hemisphere, and relatively short. Indeed, the M08 database is disproportionately composed of tree rings from North America, many of which start after 1000 CE, so that fewer than 100 records reach beyond this date. In contrast, the present collection contains 176 records out to 1000 CE, most of which are not tree-based.
I assume "records" is not the same as "series", because the Supplementary Table spreadsheet mostly lists tree ring data for North America. However, it does list 12 lake sediment series and one speleothem series. For example, NAm_065 is "Late Holocene Climate Changes in Eastern North America Estimated from Pollen Data" from https://www.cambridge.org/core/journals/quaternary-research/... .
How am I supposed to reconcile McIntyre's statement "entirely of tree rings" with the publication?
If tree rings are to be ignored, wouldn't the right question be, what does the analysis look like using only the non-tree ring data for North America?