{"id":121,"date":"2005-02-18T11:27:56","date_gmt":"2005-02-18T15:27:56","guid":{"rendered":"\/?p=121"},"modified":"2015-02-04T09:23:22","modified_gmt":"2015-02-04T14:23:22","slug":"dummies-guide-to-the-latest-hockey-stick-controversy","status":"publish","type":"post","link":"https:\/\/www.realclimate.org\/index.php\/archives\/2005\/02\/dummies-guide-to-the-latest-hockey-stick-controversy\/","title":{"rendered":"Dummies guide to the latest &#8220;Hockey Stick&#8221; controversy <lang_fr>Guide pratique pour comprendre la derni\u00e8re controverse sur la \u00ab crosse de hockey \u00bb<\/lang_fr>"},"content":{"rendered":"<div class=\"kcite-section\" kcite-section-id=\"121\">\n<p><small>by Gavin Schmidt and Caspar Amman<\/small><\/p>\n<p>Due to popular demand, we have put together a &#8216;dummies guide&#8217; which tries to describe what the actual issues are in the latest controversy, in language even our parents might understand.  A <a href=\"http:\/\/www.realclimate.org\/dummies.pdf\">pdf version<\/a> is also available. More technical descriptions of the issues can be seen  <a href=\"http:\/\/www.realclimate.org\/index.php?p=8\">here<\/a> and <a href=\"http:\/\/www.realclimate.org\/index.php?p=98\">here<\/a>. <\/p>\n<p>This guide is in two parts, the first deals with the background to the technical issues raised by <a href=\"http:\/\/www.agu.org\/pubs\/crossref\/2005\/2004GL021750.shtml\">McIntyre and McKitrick (2005)<\/a> (MM05), while the second part discusses the application of this to the original <a href=\"ftp:\/\/holocene.evsc.virginia.edu\/pub\/mann\/mbh98.pdf\">Mann, Bradley and Hughes (1998) (MBH98)<\/a> reconstruction. The wider climate science context is discussed <a href=\"http:\/\/www.realclimate.org\/index.php?p=114\">here<\/a>, and the relationship to other recent reconstructions (the &#8216;Hockey Team&#8217;) can be seen <a href=\"http:\/\/www.realclimate.org\/index.php?p=122\">here<\/a>. <\/p>\n<p>NB. All the data that were used in MBH98 are freely available for download at <a href=\"ftp:\/\/holocene.evsc.virginia.edu\/pub\/sdr\/temp\/nature\/MANNETAL98\/\">ftp:\/\/holocene.evsc.virginia.edu\/pub\/sdr\/temp\/nature\/MANNETAL98\/<\/a> (and also as supplementary data at <em><a href=\"http:\/\/www.nature.com\/nature\/journal\/v430\/n6995\/suppinfo\/nature02478.html\">Nature<\/a><\/em>) along with a thorough description of the algorithm. <\/p>\n<p><lang_fr><small>par Gavin Schmidt et Caspar Amman (traduit par Pierre Allemand)<\/small><\/p>\n<p>Suite \u00e0 la demande populaire, nous avons \u00e9crit un \u00ab guide pratique \u00bb qui essaie de d\u00e9crire quels sont les probl\u00e8mes d\u00e9battus dans la derni\u00e8re controverse sur la &#8220;crosse de hockey&#8221;, dans un langage que m\u00eame nos parents peuvent comprendre. Des descriptions techniques compl\u00e9mentaires sur ces probl\u00e8mes peuvent \u00eatre trouv\u00e9es <a href=\"http:\/\/www.realclimate.org\/index.php?p=8&#038;lp_lang_view=fr\">ici<\/a> et <a href=\"http:\/\/www.realclimate.org\/index.php?p=98&#038;lp_lang_view=fr\">ici<\/a>.<\/p>\n<p>Ce guide comporte deux parties, la premi\u00e8re concerne le fond des questions techniques soulev\u00e9es par <a href=\"http:\/\/www.agu.org\/pubs\/crossref\/2005\/2004GL021750.shtml\">McIntyre and McKitrick (2005) (MM05)<\/a>, tandis que la seconde partie traite de ses applications \u00e0 la reconstitution de <a href=\"ftp:\/\/holocene.evsc.virginia.edu\/pub\/mann\/mbh98.pdf\">Mann, Bradley et Hughes (1998) (MBH98)<\/a>. Le contexte plus large de la science du climat est discut\u00e9  <a href=\"http:\/\/www.realclimate.org\/index.php?p=114&#038;lp_lang_view=fr\">ici<\/a> et les relations avec les r\u00e9centes autres reconstitutions  (\u00ab l\u2019\u00e9quipe de hockey \u00bb) sont trait\u00e9es <a href=\"http:\/\/www.realclimate.org\/index.php?p=122&#038;lp_lang_view=fr\">ici<\/a>. <\/p>\n<p>NB. Toutes les donn\u00e9es utilis\u00e9es dans le MBH98 sont librement t\u00e9l\u00e9chargeables sur ftp:\/\/holocene.evsc.virginia.edu\/pub\/sdr\/temp\/nature\/MANNETAL98\/ (ainsi que des donn\u00e9es suppl\u00e9mentaires dans<i> <a href=\"http:\/\/www.nature.com\/nature\/journal\/v430\/n6995\/suppinfo\/nature02478.html\">Nature<\/a><\/i>) et une description compl\u00e8te de l\u2019algorithme).<br \/>\n<\/lang_fr><br \/>\n<!--more--><br \/>\nPart I: Technical issues:<\/p>\n<p>1) What is principal component analysis (PCA)?<\/p>\n<p>This is a mathematical technique that is used (among other things) to summarize the data found in a large number of noisy records so that the essential aspects can more easily seen. The most common patterns in the data are captured in a number of &#8216;principal components&#8217; which describe some percentage of the variation in the original records.  Usually only a limited number of components (&#8216;PC&#8217;s) have any statistical significance, and these can be used instead of the larger data set to give basically the same description. <\/p>\n<p>2) What do these individual components represent?<\/p>\n<p>Often the first few components represent something recognisable and physical meaningful (at least in climate data applications).  If a large part of the data set has a trend, than the mean trend may show up as one of the most important PCs. Similarly, if there is a seasonal cycle in the data, that will generally be represented by a PC. However, remember that PCs are just mathematical constructs. By themselves they say nothing about the physics of the situation. Thus, in many circumstances, physically meaningful timeseries are &#8216;distributed&#8217; over a number of PCs, each of which individually does not appear to mean much. Different methodologies or conventions can make a big difference in which pattern comes up tops. If the aim of the PCA analysis is to determine the most important pattern, then it is important to know how robust that pattern is to the methodology. However, if the idea is to more simply  summarize the larger data set, the individual ordering of the PCs is less important, and it is more crucial to make sure that as many significant PCs are included as possible. <\/p>\n<p>3) How do you know whether a PC has significant information?<\/p>\n<p><a href=\"\/images\/PCAdetails_html_m1d988d0d.png\" target=\"_blank\" title=\"PC significance (Click to enlarge)\"><img decoding=\"async\" data-src=\"\/images\/PCAdetails_html_m1d988d0d.png\" alt=\"PC significance\" align=right width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 795px; --smush-placeholder-aspect-ratio: 795\/637;\" \/><\/a> This determination is usually based on a &#8216;Monte Carlo&#8217; simulation (so-called because of the random nature of the calculations). For instance, if you take 1000 sets of random data (that have the same statistical properties as the data set in question), and you perform the PCA analysis 1000 times, there will be 1000 examples of the first PC. Each of these will explain a different amount of the variation (or variance) in the original data. When ranked in order of explained variance, the tenth one down then defines the 99% confidence level: i.e. if your real PC explains more of the variance than 99% of the random PCs, then you can say that this is significant at the 99% level. This can be done for each PC in turn. (This technique was introduced by Preisendorfer et al. (1981), and is called the Preisendorfer N-rule).<\/p>\n<p>The figure to the right gives two examples of this. Here each PC is plotted against the amount of fractional variance it explains. The blue line is the result from the random data, while the blue dots are the PC results for the real data. It is clear that at least the first two are significantly separated from the random noise line.  In the other case, there are 5 (maybe 6) red crosses that appear to be distinguishable from the red line random noise.  Note also that the first (&#8216;most important&#8217;) PC does not always explain the same amount of the original data. <\/p>\n<p>4) What do different conventions for PC analysis represent?<\/p>\n<p>Some different conventions exist regarding how the original data should be normalized. For instance, the data can be normalized to have an average of zero over the whole record, or over a selected sub-interval. The variance of the data is associated with departures from the whatever mean was selected.  So the pattern of data that shows the biggest departure from the mean will dominate the calculated PCs. If there is an <em>a priori<\/em> reason to be interested in departures from a particular mean, then this is a way to make sure that those patterns move up in the PC ordering. Changing conventions means that the explained variance of each PC can be different, the ordering can be different, and the number of significant PCs can be different.<\/p>\n<p>5) How can you tell whether you have included enough PCs?<\/p>\n<p>This is rather easy to tell. If your answer depends on the number of PCs included, then you haven&#8217;t included enough. Put another way, if the answer you get is the same as if you had used all the data without doing any PC analysis at all, then you are probably ok. However, the reason why the PC summaries are used in the first place in paleo-reconstructions is that using the full proxy set often runs into the danger of &#8216;overfitting&#8217; during the calibration period (the time period when the proxy data are trained to match the instrumental record). This can lead to a decrease in predictive skill outside of that window, which is the actual target of the reconstruction. So in summary, PC selection is a trade off: on one hand, the goal is to capture as much variability of the data as represented by the different PCs as possible (particularly if the explained variance is small), while on the other hand, you don&#8217;t want to include PCs that are not really contributing any more significant information.<\/p>\n<p>Part II: Application to the MBH98 &#8216;Hockey Stick&#8217;<\/p>\n<p>1) Where is PCA used in the MBH methodology?<\/p>\n<p>When incorporating many tree ring networks into the multi-proxy framework, it is easier to use a few leading PCs rather than 70 or so individual tree ring chronologies from a particular region. The trees are often very closely located and so it makes sense to summarize the general information they all contain in relation to the  large-scale patterns of variability. The relevant signal for the climate reconstruction is the signal that the trees have in common, not each individual series. In MBH98, the North American tree ring series were treated like this. There are a number of other places in the overall methodology where some form of PCA was used, but they are not relevant to this particular controversy.<\/p>\n<p>2) What is the point of contention in MM05?<\/p>\n<p>MM05 contend that the particular PC convention used in MBH98 in dealing with the N. American tree rings selects for the &#8216;hockey stick&#8217; shape and that the final reconstruction result is simply an artifact of this convention.<\/p>\n<p>3) What convention was used in MBH98?<\/p>\n<p>MBH98 were particularly interested in whether the tree ring data showed significant differences from the 20th century calibration period, and therefore normalized the data so that the mean over this period was zero. As discussed above, this will emphasize records that have the biggest differences from that period (either positive of negative). Since the underlying data have a &#8216;hockey stick&#8217;-like shape, it is therefore not surprising that the most important PC found using this convention resembles the &#8216;hockey stick&#8217;. There are actual two significant PCs found using this convention, and both were incorporated into the full reconstruction.<\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m2a00a61d.png\" target=\"_blank\" title=\"PC1 vs PC4 (Click to enlarge)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m2a00a61d.png\" alt=\"PC1 vs PC4\" align=right width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 948px; --smush-placeholder-aspect-ratio: 948\/713;\" \/><\/a><\/p>\n<p>4) Does using a different convention change the answer?<\/p>\n<p>As discussed above, a different convention (MM05 suggest one that has zero mean over the whole record) will change the ordering, significance and number of important PCs.  In this case, the number of significant PCs increases to 5 (maybe 6) from 2 originally. This is the difference between the blue points (MBH98 convention) and the red crosses (MM05 convention) in the first figure. Also PC1 in the MBH98 convention moves down to PC4 in the MM05 convention. This is illustrated in the figure on the right, the red curve is the original PC1 and the blue curve is MM05 PC4 (adjusted to have same variance and mean). But as we stated above, the underlying data has a hockey stick structure, and so in either case the &#8216;hockey stick&#8217;-like PC explains a significant part of the variance. Therefore, using the MM05 convention, more PCs need to be included to capture the significant information contained in the tree ring network. <\/p>\n<p>This figure shows the difference in the final result whether you use the original convention and 2 PCs (blue) and the MM05 convention with 5 PCs (red). The MM05-based reconstruction is slightly less skilful when judged over the 19th century validation period but is otherwise very similar. In fact any calibration convention will lead to approximately the same answer as long as the PC decomposition is done properly and one determines how many PCs are needed to retain the primary information in the original data.<\/p>\n<p><a href=\"\/images\/\/FalseClaimsMcIntyreMcKitrick_html_m7e711aa.png\" target=\"_blank\" title=\"different conventions (Click to enlarge)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m7e711aa.png\" alt=\"different conventions\" width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 908px; --smush-placeholder-aspect-ratio: 908\/722;\" \/><\/a><\/p>\n<p>5) What happens if you just use all the data and skip the whole PCA step?<\/p>\n<p>This is a key point. If the PCs being used were inadequate in characterizing the underlying data, then the answer you get using all of the data will be significantly different. If, on the other hand, enough PCs were used, the answer should be essentially unchanged. This is shown in the figure below. The reconstruction using all the data is in yellow (the green line is the same thing but with  the &#8216;St-Anne River&#8217; tree ring chronology taken out). The blue line is the original reconstruction, and as you can see the correspondence between them is high.  The validation is slightly worse, illustrating the trade-off mentioned above i.e. when using all of the data, over-fitting during the calibration period (due to the increase number of degrees of freedom) leads to a slight loss of predictability in the validation step. <\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_7955bc86.png\" target=\"_blank\" title=\"No PCA comparison (Click to enlarge)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_7955bc86.png\" alt=\"No PCA comparison\"  width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1270px; --smush-placeholder-aspect-ratio: 1270\/975;\" \/><\/a><\/p>\n<p>6) So how do MM05 conclude that this small detail changes the answer?<\/p>\n<p>MM05 claim that the reconstruction using only the first 2 PCs with their convention is significantly different to MBH98. Since PC 3,4 and 5 (at least) are also significant they are leaving out good data. It is mathematically wrong to retain the same number of PCs if the convention of standardization is changed. In this case, it causes a loss of information that is very easily demonstrated. Firstly, by showing that any such results do not resemble the results from using all data, and by checking the validation of the reconstruction for the 19th century.  The MM version of the reconstruction can be matched by simply removing the N. American tree ring data along with the &#8216;St Anne River&#8217; Northern treeline series from the reconstruction (shown in yellow below). Compare this curve with the ones shown above. <\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_32969d82.png\" target=\"_blank\" title=\"No N. American tree rings (Click to enlarge)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_32969d82.png\" alt=\"No N. American tree rings\"  width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1265px; --smush-placeholder-aspect-ratio: 1265\/996;\" \/><\/a><\/p>\n<p>As you might expect, throwing out data also worsens the validation statistics, as can be seen by eye when comparing the reconstructions over the 19th century validation interval. Compare the green line in the figure below to the instrumental data in red. To their credit, MM05 acknowledge that their alternate 15th century reconstruction has no skill. <\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m3db92977.png\" target=\"_blank\" title=\"validation period (Click to enlarge)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m3db92977.png\" alt=\"validation period\"  width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1289px; --smush-placeholder-aspect-ratio: 1289\/976;\" \/><\/a><\/p>\n<p>7) Basically then the MM05 criticism is simply about whether selected N. American tree rings should have been included, not that there was a mathematical flaw?<\/p>\n<p>Yes. Their argument since the beginning has essentially <em>not<\/em> been about methodological issues at all, but about &#8216;source data&#8217; issues. Particular concerns with the &#8220;bristlecone pine&#8221; data were addressed in the followup paper MBH99 but the fact remains that including these data improves the statistical validation over the 19th Century period and they therefore should be included. <\/p>\n<p><a href=\"http:\/\/en.wikipedia.org\/wiki\/Image:1000_Year_Temperature_Comparison.png\" target=\"_blank\" title=\"Hockey Team *used under GFDL license (Click to enlarge)\"><img decoding=\"async\" data-src=\"http:\/\/www.wmconnolley.org.uk\/1000_Year_Temperature_Comparison.png\" alt=\"Hockey Team *used under GFDL license\" align=right width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" \/><\/a><\/p>\n<p>8) So does this all matter?<\/p>\n<p>No. If you use the MM05 convention and include all the significant PCs, you get the same answer. If you don&#8217;t use any PCA at all, you get the same answer. If you use a completely different methodology (i.e. Rutherford et al, 2005), you get basically the same answer. Only if you remove significant portions of the data do you get a different (and worse) answer.<\/p>\n<p>9) Was MBH98 the final word on the climate of last millennium?<\/p>\n<p>Not at all. There has been significant progress on many aspects of climate reconstructions since MBH98. Firstly, there are more and better quality proxy data available. There are new methodologies such as described in <a href=\"http:\/\/www.realclimate.org\/RuthetalJClim2004.pdf\">Rutherford et al (2005)<\/a> or <a href=\"http:\/\/www.realclimate.org\/index.php?p=122\">Moberg et al (2005)<\/a> that address recognised problems with incomplete data series and the challenge of incorporating lower resolution data into the mix. Progress is likely to continue on all these fronts. As of now, all of the &#8216;Hockey Team&#8217; reconstructions (shown left) agree that the late 20th century is anomalous in the context of last millennium, and possibly the last two millennia.<br \/>\n<lang_fr><br \/>\n1\u00e8re partie : Questions Techniques :<\/p>\n<p>1)   Quelle est la signification de \u201cl\u2019analyse en composantes principales \u201c (ACP) ?<\/p>\n<p>C\u2019est une technique math\u00e9matique utilis\u00e9e (entre autres) pour condenser les donn\u00e9es fournies par une grande s\u00e9rie de chiffres dispers\u00e9s de fa\u00e7on \u00e0 ce que les aspects essentiels apparaissent plus facilement. Les mod\u00e8les les plus courants retrouv\u00e9s dans les donn\u00e9es sont regroup\u00e9s en un certain nombre de \u2018composantes principales\u2019 qui d\u00e9crivent une certaine fraction de la variation des donn\u00e9es d\u2019origine. Normalement, il existe un nombre limit\u00e9 de composantes dites principales (CP) poss\u00e9dant une certaine signification statistique et pouvant \u00eatre utilis\u00e9s \u00e0 la place de l\u2019ensemble des chiffres pour aboutir fondamentalement \u00e0 la m\u00eame description.<\/p>\n<p>2)   Que repr\u00e9sentent ces composantes individuelles ?<\/p>\n<p>Souvent, les quelques premieres composantes repr\u00e9sentent quelque chose de reconnaissable et de physiquement significatif (du moins dans le domaine climatique). Si une fraction importante de ces donn\u00e9es montre une tendance, alors, la tendance moyenne peut r\u00e9v\u00e9ler une des composantes principales (CP) les plus importantes. De la m\u00eame mani\u00e8re, s\u2019il y a un cycle saisonnier dans les chiffres, celui-ci sera repr\u00e9sent\u00e9 par une CP. N\u00e9anmoins, il faut se souvenir que les CP sont de simples repr\u00e9sentations math\u00e9matiques. En soi, elles n\u2019ont aucune signification physique. Ainsi, dans de nombreux cas, des s\u00e9ries temporelles physiquement significatives sont \u201cdistribu\u00e9es\u201d entre un certain nombre de CP qui n\u2019apparaissent pas tr\u00e8s significatives prises individuellement. Le choix de diff\u00e9rentes m\u00e9thodologies ou conventions peut cr\u00e9er d\u2019importantes diff\u00e9rences quant au modele structural (\u2018pattern\u2019) qui appara\u00eet pr\u00e9pond\u00e9rant. Si l\u2019objectif de l\u2019analyse ACP est de d\u00e9terminer le mod\u00e8le structural le plus important, alors il est important de conna\u00eetre la solidit\u00e9 de ce mod\u00e8le structural vis \u00e0 vis de la m\u00e9thodologie. N\u00e9anmoins, si l\u2019id\u00e9e est, plus simplement, de condenser un grand nombre de chiffres, l\u2019ordre d\u2019importance des CP devient moins important, et il est plus crucial de s\u2019assurer que le plus grand nombre possible de CP significatives a \u00e9t\u00e9 pris en compte. <\/p>\n<p>3)   Comment savez-vous qu\u2019une CP apporte des informations significatives?<\/p>\n<p> <a href=\"\/images\/PCAdetails_html_m1d988d0d.png\" target=\"_blank\" title=\"validit\u00e9 des CPs (Cliquez pour agrandir)\"><img decoding=\"async\" data-src=\"\/images\/PCAdetails_html_m1d988d0d.png\" alt=\"validit\u00e9 des CPs\" align=right width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 795px; --smush-placeholder-aspect-ratio: 795\/637;\" \/><\/a><br \/>\nCette d\u00e9termination est g\u00e9n\u00e9ralement fond\u00e9e sur une \u00absimulation de Monte Carlo \u00bb (ainsi appel\u00e9e en raison de la nature probabiliste des calculs). Par exemple, si vous prenez 1000 groupes de chiffres au hasard (qui auront les m\u00eames propri\u00e9t\u00e9s statistiques que le groupe de chiffres en question), et que vous effectuez 1000 fois l\u2019analyse ACP, vous obtiendrez 1000 diff\u00e9rentes premieres CP. Chacun d\u2019entre eux expliquera une quantit\u00e9 diff\u00e9rente de variation (ou variance) dans les chiffres d\u2019origine. Ordonn\u00e9s par degr\u00e9 d\u2019explication de variance, le d\u00e9cile inf\u00e9rieur d\u00e9fini le niveau de confiance 99 % : c\u2019est \u00e0 dire que si votre CP r\u00e9elle explique plus de variance que 99% du CP al\u00e9atoire, alors vous pouvez dire qu\u2019il est significatif \u00e0 99 %. Ceci peut \u00eatre effectu\u00e9 successivement pour chacune des CP. (Cette technique a \u00e9t\u00e9 introduite par Preisendorfer et ses coll\u00e8gues (1981), et est appel\u00e9e la r\u00e8gle normative de Preisendorfer).<br \/>\nLa figure de droite donne deux exemples de cela. Dans ce graphique, chaque CP est positionn\u00e9e en fonction de la fraction de variance qu\u2019elle repr\u00e9sente. La ligne bleue correspond \u00e0 une s\u00e9rie al\u00e9atoire de chiffres, alors que les points bleus sont le r\u00e9sultat de CP sur des chiffres r\u00e9els. Il appara\u00eet clairement qu\u2019au moins les 2 premiers points sont significativement s\u00e9par\u00e9s de la ligne de points al\u00e9atoires. Dans l\u2019autre cas, il y a 5 (peut-\u00eatre 6) croix rouges qui apparaissent discernables de la ligne rouge des points al\u00e9atoires. Notez aussi que la premiere (\u2018la plus importante\u2019) CP n\u2019explique pas toujours la m\u00eame quantit\u00e9 de chiffres originaux.<\/p>\n<p>4)   Que repr\u00e9sentent les diff\u00e9rentes conventions pour l\u2019analyse des CP ?<\/p>\n<p>Il existe quelques diff\u00e9rentes conventions concernant la fa\u00e7on dont les donn\u00e9es originelles doivent \u00eatre normalis\u00e9es. Par exemple, les chiffres peuvent \u00eatre normalis\u00e9s pour que leur moyenne soit de z\u00e9ro sur l\u2019ensemble de la s\u00e9rie, ou alors sur un sous-intervalle choisi. La variance est associ\u00e9e aux \u00e9carts par rapport \u00e0 la moyenne quelle que soit la fa\u00e7on dont celle-ci a \u00e9t\u00e9 choisie.<br \/>\nAinsi, le mod\u00e8le structural des donn\u00e9es qui montre les plus grands \u00e9carts par rapport \u00e0 la moyenne dominera dans le calcul des CP. S\u2019il y a une raison a priori de s\u2019int\u00e9resser aux \u00e9carts par rapport \u00e0 une certaine moyenne, ceci est un moyen de s\u2019assurer que les mod\u00e8les structuraux r\u00e9sultants seront plac\u00e9s en t\u00eate dans l\u2019ordre des CP. Changer les conventions signifie que la quantit\u00e9 de variance expliqu\u00e9e peut \u00eatre diff\u00e9rente pour chacun des CP, ainsi que leur ordre d\u2019importance et le nombre des CP significatives.<\/p>\n<p>5)   Comment pouvez-vous dire que vous avez pris en compte assez de CP ?<\/p>\n<p>C\u2019est assez facile \u00e0 dire. Si votre r\u00e9ponse d\u00e9pend du nombre de CP introduites, cela signifie qu\u2019il n\u2019y en a pas assez. Dit d\u2019une autre fa\u00e7on, si la r\u00e9ponse que vous obtenez est la m\u00eame que si vous aviez utilis\u00e9 toutes les donn\u00e9es sans faire aucune analyse de CP, alors vous \u00eates probablement au bon nombre. N\u00e9anmoins, la raison pour laquelle les condensations de donn\u00e9es par CP sont utilis\u00e9es en premier lieu dans les pal\u00e9o reconstitutions est que l\u2019utilisation de la totalit\u00e9 des indicateurs introduit le danger de trop bien coller aux chiffres de la p\u00e9riode de calibrage (la p\u00e9riode durant laquelle les indicateurs sont soumis \u00e0 correspondre aux mesures instrumentales). Cela peut conduire \u00e0 une d\u00e9ficience dans les capacit\u00e9s pr\u00e9dictives en dehors de cette fen\u00eatre, alors que c\u2019est pr\u00e9cis\u00e9ment l\u2019objectif de la reconstitution. Pour r\u00e9sumer, la s\u00e9lection des CP est un compromis : d\u2019un cot\u00e9, l\u2019objectif est de capturer autant de variabilit\u00e9 des donn\u00e9es que possible repr\u00e9sent\u00e9e par les diff\u00e9rents CP. De l\u2019autre, il ne faut pas introduire de CP qui n\u2019apporte pas r\u00e9ellement d\u2019information significative suppl\u00e9mentaire. <\/p>\n<p>Partie II : Application \u00e0 l\u2019\u00e9quipe de hockey (MBH98)<\/p>\n<p>1)   O\u00f9 donc est utilis\u00e9e l\u2019Analyse des Composantes principales (ACP) dans la m\u00e9thodologie MBH ?<\/p>\n<p>Quand on introduit un ensemble d\u2019anneaux de croissance dans un syst\u00e8me multi-marqueurs, il est facile d\u2019utiliser quelques CP pr\u00e9pond\u00e9rantes, plut\u00f4t que quelques 70 chronologies individuelles d\u2019anneaux de croissance en provenance d\u2019une r\u00e9gion particuli\u00e8re. Les arbres sont souvent tr\u00e8s voisins, et donc, il para\u00eet logique de condenser les informations g\u00e9n\u00e9rales qu\u2019ils contiennent d\u2019apr\u00e8s des mod\u00e8les de variabilit\u00e9 plus larges.<br \/>\nLe signal significatif pour la reconstitution du climat est le signal commun \u00e0 tous les arbres, et non chacune des s\u00e9ries individuelles. Dans MBH98, les s\u00e9ries d\u2019anneaux de croissance des arbres nord am\u00e9ricains ont \u00e9t\u00e9 trait\u00e9es de cette fa\u00e7on. Dans un certain nombre d\u2019autres endroits o\u00f9 la m\u00e9thodologie g\u00e9n\u00e9rale a \u00e9t\u00e9 appliqu\u00e9e, certaines formes d\u2019ACP ont \u00e9t\u00e9 utilis\u00e9es, mais elles ne rel\u00e8vent pas de cette controverse en particulier. <\/p>\n<p>2)   Quel est le point controvers\u00e9 dans MMH98 ?<\/p>\n<p>MM05 conteste le fait que la convention particuli\u00e8re utilis\u00e9e pour le choix des CP par MBH98 s\u00e9lectionne les anneaux de croissance d\u2019arbres nord am\u00e9ricains pour obtenir la forme en \u00ab crosse de hockey \u00bb, et que donc la reconstitution finale est simplement un art\u00e9fact de cette convention.<\/p>\n<p>3)   Quelle convention a-t-elle \u00e9t\u00e9 utilis\u00e9e dans MBH98 ?<\/p>\n<p>MBH98 se sont particuli\u00e8rement int\u00e9ress\u00e9s \u00e0 savoir si les anneaux de croissance mettaient en \u00e9vidence des diff\u00e9rences significatives par rapport \u00e0 la p\u00e9riode de calibrage au 20\u00e8me si\u00e8cle, et en cons\u00e9quence, ont normalis\u00e9 les donn\u00e9es pour que la moyenne dans cette p\u00e9riode soit \u00e9gale \u00e0 z\u00e9ro. Comme expos\u00e9 plus haut, cela augmente l\u2019importance des valeurs qui pr\u00e9sente les plus grandes diff\u00e9rences avec cette p\u00e9riode (aussi bien n\u00e9gatives que positives). Comme la courbe de r\u00e9f\u00e9rence est en forme de \u00ab crosse de hockey \u00bb, il n\u2019est pas \u00e9tonnant que les CP les plus importantes trouv\u00e9es en utilisant cette convention soient eux-m\u00eames en forme de \u00ab crosse de hockey \u00bb. On trouve, en fait deux CP significatifs en utilisant cette convention, et ils ont \u00e9t\u00e9 incorpor\u00e9s tous les deux dans la reconstitution compl\u00e8te.<\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m2a00a61d.png\" target=\"_blank\" title=\"PC1 vs PC4 (Cliquez pour agrandir)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m2a00a61d.png\" alt=\"PC1 vs PC4\" align=right width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 948px; --smush-placeholder-aspect-ratio: 948\/713;\" \/><\/a><\/p>\n<p>4)   Est-ce que le fait de changer de convention change la r\u00e9ponse ?<\/p>\n<p>Comme discut\u00e9 plus haut, une convention diff\u00e9rente (MM05 sugg\u00e8rent celle de fixer \u00e0 z\u00e9ro la moyenne sur l\u2019ensemble des donn\u00e9es) changera l\u2019ordre, le degr\u00e9 de signification, et le nombre d\u2019importants CP. Dans ce cas, le nombre de CP significatifs monte \u00e0 5 (peut-\u00eatre 6) au lieu de 2 \u00e0 l\u2019origine. Voici la diff\u00e9rence entre les points bleus (convention de MBH98) et les croix rouges (convention de MM05) dans le premier graphique. Le CP N\u00b01 dans la convention de MBH98 descend au CP N\u00b0 4 dans la convention de MM05.<br \/>\nCeci est illustr\u00e9 par la figure de droite, la courbe rouge repr\u00e9sente le CP N\u00b0 1 d\u2019origine et la courbe bleue le CP n\u00b0 4 de MM05 (ajust\u00e9es pour avoir la m\u00eame variance et la m\u00eame moyenne).<\/p>\n<p>Mais, comme nous l\u2019avons fait remarquer plus haut, les donn\u00e9es sous-jacentes ont une structure en forme de crosse de hockey, si bien que dans les deux cas, le CP en forme de crosse de hockey explique une partie significative de la variance. En cons\u00e9quence, lors de l\u2019utilisation de la convention MM05, on doit utiliser un nombre plus grand de CP afin de saisir les informations significatives contenues dans l\u2019ensemble des anneaux de croissance.<\/p>\n<p>Cette figure montre la diff\u00e9rence sur le r\u00e9sultat final selon que l\u2019on utilise la convention d\u2019origine avec 2 CP (bleu) ou la convention MM05 avec 5 CP (rouge).<br \/>\nLa reconstitution type MM05 est l\u00e9g\u00e8rement moins performante si on l\u2019examine sur la p\u00e9riode de validation du 19\u00e8me si\u00e8cle, mais se trouve par ailleurs tr\u00e8s similaire.. En fait, n\u2019importe quelle convention de calibrage conduira approximativement \u00e0 la m\u00eame r\u00e9ponse, \u00e0 condition d\u2019effectuer convenablement la d\u00e9composition en CP, et que l\u2019on d\u00e9termine combien de CP sont n\u00e9cessaires pour retenir les informations fondamentales contenues dans les donn\u00e9es d\u2019origine. <\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m7e711aa.png\" target=\"_blank\" title=\"different conventions (Cliquez pour agrandir)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m7e711aa.png\" alt=\"different conventions\" width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 908px; --smush-placeholder-aspect-ratio: 908\/722;\" \/><\/a><br \/>\n5)   Que se passe-t-il si vous utilisez simplement tous les chiffres et sautez enti\u00e8rement l\u2019\u00e9tape ACP ?<\/p>\n<p>C\u2019est est un point cl\u00e9. Si les CP utilis\u00e9s ne sont pas adapt\u00e9s \u00e0 la caract\u00e9risation des donn\u00e9es sous jacentes, alors la r\u00e9ponse obtenue en utilisant toutes les donn\u00e9es sera significativement diff\u00e9rente. D\u2019un autre cot\u00e9, si on utilise un nombre suffisant de CP, la r\u00e9ponse sera essentiellement identique.<br \/>\nCette figure le montre.<br \/>\nLa reconstitution utilisant toutes les donn\u00e9es est en jaune. (La courbe verte repr\u00e9sente la m\u00eame chose, mais sans les donn\u00e9es des anneaux de croissance de la \u00ab Rivi\u00e8re Ste Anne \u00bb. La courbe bleue repr\u00e9sente la reconstitution originelle, et, comme on peut le constater, la correspondance entre les courbes est grande.<\/p>\n<p>La validation est l\u00e9g\u00e8rement moins bonne, ce qui illustre le compromis mentionn\u00e9 plus haut : lorsqu\u2019on utilise toutes les donn\u00e9es, la prise en compte trop pr\u00e9cise de toutes les donn\u00e9es pendant la p\u00e9riode de calibrage ( \u00e0 cause du nombre plus important de degr\u00e9s de libert\u00e9) conduit \u00e0 une l\u00e9g\u00e8re perte de pr\u00e9dictibilit\u00e9 au moment de la validation.<\/p>\n<p> <a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_7955bc86.png\" target=\"_blank\" title=\"Pas de PCA (Cliquez pour agrandir)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_7955bc86.png\" alt=\"No PCA comparison\"  width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1270px; --smush-placeholder-aspect-ratio: 1270\/975;\" \/><\/a><\/p>\n<p>6)   Donc, comment MM05 peuvent-il conclure que ce petit d\u00e9tail change la r\u00e9ponse ?<\/p>\n<p>MM05 affirment que la reconstitution utilisant seulement les 2 premieres CP avec leur convention est significativement diff\u00e9rente de celle de MBH98. Comme les CP n\u00b0 3, 4 et 5 (au moins) sont aussi significatives, des donn\u00e9es int\u00e9ressantes ne sont pas prises en compte. Il est math\u00e9matiquement incorrect de retenir le m\u00eame nombre de CP si la convention de standardisation est chang\u00e9e.<\/p>\n<p>Dans ce cas, la perte d\u2019informations est facilement d\u00e9montrable. D\u2019abord en montrant que de tels r\u00e9sultats ne ressemblent pas aux r\u00e9sultats qui utilisent toutes les donn\u00e9es, et en v\u00e9rifiant la validit\u00e9 de la reconstitution pour le 19\u00e8me si\u00e8cle.<br \/>\nLa version MM de la reconstitution peut \u00eatre reproduite en \u00e9liminant simplement les donn\u00e9es des anneaux de croissance nord am\u00e9ricains en m\u00eame temps que la s\u00e9rie des anneaux nordiques de la \u00ab Rivi\u00e8re Ste Anne \u00bb de la reconstitution (courbe jaune). Comparez cette courbe avec les courbes montr\u00e9es plus haut).<\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_32969d82.png\" target=\"_blank\" title=\"Pas des anneaus d'arbes Americanes (Cliquez pour agrandir)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_32969d82.png\" alt=\"No N. American tree rings\"  width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1265px; --smush-placeholder-aspect-ratio: 1265\/996;\" \/><\/a><\/p>\n<p>Comme vous pouviez vous y attendre, rejeter des donn\u00e9es d\u00e9t\u00e9riore aussi les statistiques de validation comme cela appara\u00eet visuellement en comparant les reconstitutions sur l\u2019intervalle de validation du 19\u00e8me si\u00e8cle. Comparez la courbe verte de la figure suivante aux mesures instrumentales en rouge. A leur cr\u00e9dit, MM05 reconnaissent que leur reconstitution alternative du 15\u00e8me si\u00e8cle n\u2019a pas beaucoup de valeur.<\/p>\n<p><a href=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m3db92977.png\" target=\"_blank\" title=\"periode de validation (Cliquez pour agrandir)\"><img decoding=\"async\" data-src=\"\/images\/FalseClaimsMcIntyreMcKitrick_html_m3db92977.png\" alt=\"validation period\"  width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1289px; --smush-placeholder-aspect-ratio: 1289\/976;\" \/><\/a><\/p>\n<p>7)   Dans le fond, alors, la critique de MM05 porte simplement sur le fait que MBH98 a utilis\u00e9 une s\u00e9lection d\u2019anneaux de croissance d\u2019arbres nord am\u00e9ricains, et pas sur une erreur math\u00e9matique ?<\/p>\n<p>Oui. Leur argument depuis le d\u00e9but porte essentiellement non pas du tout sur des questions m\u00e9thodologiques, mais sur la question de l\u2019origine des donn\u00e9es. Des probl\u00e8mes particuliers concernant les pins bristlecones (<em>Pinus longaeva<\/em>) ont \u00e9t\u00e9 signal\u00e9s dans l\u2019article suivant MBH99, mais le fait reste que la prise en compte de ces donn\u00e9es am\u00e9liore la validation statistique sur la p\u00e9riode du 19\u00e8me si\u00e8cle et qu\u2019elles devraient \u00eatre prises en compte.<\/p>\n<p> <a href=\"http:\/\/en.wikipedia.org\/wiki\/Image:1000_Year_Temperature_Comparison.png\" target=\"_blank\"><img decoding=\"async\" data-src=\"http:\/\/www.wmconnolley.org.uk\/1000_Year_Temperature_Comparison.png\" alt=\"Hockey Team *used under GFDL license\" align=right width=300 src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" \/><\/a><br \/>\n8)   Est-ce que cela est important ?<\/p>\n<p>Non. Si vous utilisez la convention MM05 et que vous prenez en compte tous les CP significatifs, vous obtenez la m\u00eame r\u00e9ponse. Si vous n\u2019utilisez aucune PCA, vous obtenez le m\u00eame r\u00e9sultat. Si vous utilisez une m\u00e9thodologie compl\u00e8tement diff\u00e9rente, (Rutherford et al. 2005) vous obtenez fondamentalement la m\u00eame r\u00e9ponse. C\u2019est seulement si vous \u00e9liminez des portions significatives de donn\u00e9es que vous obtenez une r\u00e9ponse diff\u00e9rente (et plus mauvaise). <\/p>\n<p>9)   Est-ce que MBH98 repr\u00e9sente le dernier mot sur le climat du dernier mill\u00e9naire ?<br \/>\nPas du tout. Il y a eu des progr\u00e8s significatifs sur de nombreux aspects de la reconstitution du climat depuis MBH98. D\u2019abord, il existe des indicateurs plus nombreux, et de meilleure qualit\u00e9. Il y a de nouvelles m\u00e9thodologies comme celles d\u00e9crites dans <a href=\"http:\/\/www.realclimate.org\/RuthetalJClim2004.pdf\">Rutherford et al (2005)<\/a> ou <a href=\"http:\/\/www.realclimate.org\/index.php?p=122&#038;lp_lang_view=fr\">Moberg et al (2005)<\/a> qui mettent en avant des probl\u00e8mes reconnus de s\u00e9ries de donn\u00e9es incompl\u00e8tes, ainsi que la question de l\u2019introduction dans l\u2019ensemble des chiffres pris en compte, de donn\u00e9es \u00e0 r\u00e9solution plus basse. On constate des progr\u00e8s dans pratiquement tous ces domaines. Pour le moment, toutes les reconstitutions de \u00ab l\u2019\u00e9quipe de hockey \u00bb (voir \u00e0 gauche) sont d\u2019accord sur le fait que le 20\u00e8me si\u00e8cle se pr\u00e9sente comme une anomalie dans le contexte du dernier mill\u00e9naire, et peut-\u00eatre des deux derniers.<br \/>\n<\/lang_fr><\/p>\n<!-- kcite active, but no citations found -->\n<\/div> <!-- kcite-section 121 -->","protected":false},"excerpt":{"rendered":"<p>by Gavin Schmidt and Caspar Amman Due to popular demand, we have put together a &#8216;dummies guide&#8217; which tries to describe what the actual issues are in the latest controversy, in language even our parents might understand. A pdf version is also available. More technical descriptions of the issues can be seen here and here. [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_exactmetrics_skip_tracking":false,"_exactmetrics_sitenote_active":false,"_exactmetrics_sitenote_note":"","_exactmetrics_sitenote_category":0,"_genesis_hide_title":false,"_genesis_hide_breadcrumbs":false,"_genesis_hide_singular_image":false,"_genesis_hide_footer_widgets":false,"_genesis_custom_body_class":"","_genesis_custom_post_class":"","_genesis_layout":"","footnotes":""},"categories":[1,13,2],"tags":[],"class_list":{"0":"post-121","1":"post","2":"type-post","3":"status-publish","4":"format-standard","6":"category-climate-science","7":"category-faq","8":"category-paleoclimate","9":"entry"},"aioseo_notices":[],"post_mailing_queue_ids":[],"_links":{"self":[{"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/posts\/121","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/comments?post=121"}],"version-history":[{"count":2,"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/posts\/121\/revisions"}],"predecessor-version":[{"id":18081,"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/posts\/121\/revisions\/18081"}],"wp:attachment":[{"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/media?parent=121"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/categories?post=121"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.realclimate.org\/index.php\/wp-json\/wp\/v2\/tags?post=121"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}