Righto, back again.
brazil84, it seems to be pretty apparent that you don’t understand at least a good half of what everyone else is writing and are probably just mentally skipping it, perhaps without even realising. So really what we need to do first is figure out where your understanding is breaking down.
First question:
Would you say that as the average global temperature rises and falls that every location on earth will rise and fall in temperature by the same amount?
I would say that it depends on the trend. For example, if there is an upward trend that lasts 5 years, a 100 year moving average probably won’t pick it up. On the other hand, a long term trend might be easier to see against year to year fluctuation if you use a longer moving average.
But why won’t you answer my question? It’s a very simple question:
Let’s define an “elbow” as a point where there is a large change in the slope of the graph. All things being equal, a graph with a 40 year smooth will have equal or fewer elbows than a graph with a 5 year smooth. Right?
Why don’t you just answer my question? It might be that we’re on the same page here.
It depends on the poster. Anyway, perhaps you are the one who doesn’t understand.
I would say probably not.
Well, I’m not sure if its been mentioned or not, because my eyes have glazed over at times reading the thread, but I would strongly recommend that anyone interested in the topic read this thread from beginning to end.
Then you can come back and tell us if you found any factual errors in the presentation.
The hockeystick is deader than dead, and all of Michael Mann’s friends and supporters can’t put Humpty back together again.
w.
Yes…I can find some. For example, McIntyre cherrypicks quotes by quoting this long section from the report:
while ignoring [section:
(bolding added)
So, in other words, yes, this method is a potentially dangerous one to use but, no, it did not in fact make a difference in this case. This would hardly be the first case when a scientist has gotten the right answer with the wrong (or a not completely robust) method. One reason this happens is that most good scientists will actually do the calculation a few different ways, so while he may write up this particular technique, he may well have verified the results with another technique.
The NAS panel do agree with McIntyre and McKitrick more in regard to the issue of robustness with respect to the proxy choices but that is a different issue…and the fact that “the Mann et al. (1999) reconstruction that uses this particular principal component analysis technique is strongly dependent on data from the Great Basin region in the western United States” was in fact something that [url=http://wdc.obs-mip.fr/paleo/pubs/millennium-camera.pdf]was already pointed out by Mann et al. (1999) themselves](]this[/url):
(ITRDB stands for the International Tree Ring DataBank [all from North America] and it is noted elsewhere the Principle Component (PC) #1 is dominated by the western U.S. region.)
And, elsewhere in that paper, Mann et al. say:
(bolding added)
The reason why this statement is so interesting is the McIntyre and McKitrick made such a big deal about finding a directory on Mann’s website that had the word “CENSORED” or something like that in the directory name that showed how the reconstruction looks in the absence of that data. This was their smoking-gun proof that Mann et al. were essentially perpetrating fraud, i.e., intentionally covering up negative results. Of course, if you are really going to censor a result, there are probably better ways of doing than reporting it in a paper published in Geophysical Review Letters!!
I could have sworn this had been brought up before, either in this thread or the BBQ Pit thread. Are rehashing the same points again?
I think that’s the real question for this thread. The pro-hockey stick people seem to be saying that even if Mann did things he shouldn’t have done, if you do things right you still get a hockey stick.
Of course, one then needs to define “hockey stick,” which I tried to do earlier:
The real question, it seems to me, is whether these conclusions survive scrutiny.
jshore, you do not dispute the truth of McIntyre’s quotes. I note you say he cherry picks by not including this quote:
Unfortunately, you are cherry picking here. All the quote shows is that
The method Mann used is in fact wrong. You did notice that part?
If you include the discredited bristlecones, you can get the hockeystick shape by using OLS or any one of a number of matters.
If that is supposed to invalidate the numerous problems with the Mann corpus of claims, I fear you’re betting on the wrong horse. Yes, you are correct, they said you can use any number of methods to get a hockeystick from bad data … is that intended to impress us?
w.
Thanks for your reply, intention.
I don’t understand your statement at all. In what way did I cherry pick? McIntyre purposely withheld information from his readers (and we know he couldn’t possibly have missed the remark that I quoted since in another quotation, he quoted both a piece before it and a piece immediately after it with “…” in between). I gave you the full quote in that section on that subject, as well as the full quote that McIntyre had given, and then even referred to the sentences following the quote that I gave which were more in agreement with McIntyre and McKitrick on a tangentially-related but different point.
Did you miss the part where I said:
First of all, the bristlecones are not “discredited”. There are certain concerns about that data, as there is with lots of the different temperature proxy data. As I noted, it was Mann et al. themselves that noted in their 1999 paper that the results (and the quality of the fit they could obtain to the instrumental temperature record during the period of overlap) was strongly dependent on data from that region, noted some concerns with that data and how they had tried to address them, and noted that “Clearly, a more widespread network of quality millennial proxy climate indicators will be required for more confident inferences.”
Again, you are hampered by some binary notion that data is either perfect or totally discredited. In fact, in real science at the cutting-edge of any field, there are always limitations in the quality of the data and needs for improvements. And, that is why the IPCC in both the TAR and AR4 chose to assess the probability that the Northern Hemisphere temperatures were warmer in the latter half of the twentieth century than in any previous century in the last 1000 - 1300 years as a “likely” statement, i.e., >66% of being true.
So, basically, what McIntyre and McKitrick have shown is:
(1) That the principle components technique that Mann et al. presented is not robust and can be dangerous to use in the sense that it can lead to hockey-stick shapes from data that does not in fact contain it, although that this did not happen in this case since other ways of analyzing the data give the same result. This result is certainly interesting and relevant to those who actually might want to use the technique in the future but turns out to be irrelevant to the results actually presented. [I think Mann et al. have further technical arguments about how the hockey stick shapes that McIntyre and McKitrick came up with from synthetic data would not have actually passed their significance tests but this gets into arguments about the best significance tests to use and so forth that frankly I haven’t taken the time to understand.]
(2) The sensitivity to certain data in the reconstruction that was already pointed out by Mann et al. in their 1999 paper. I suppose one can give M&M a little more credit here for sort of expanding on this issue…which does seem like an important factor for future data improvements, as was in fact noted by Mann et al. in 1999.
So, of these two points, one is irrelevant unless you are an actual practitioner planning to apply this technique to your data and the other is relevant but already noted by Mann et al. and thus presumably pretty well-understood by the scientists in the field and incorporated in their assessment of the uncertainties regarding the results.
Second question:
If I was to ask 100 people to graph the fall of a marble (distance/time) in a vacuum, should I expect 100 different looking graphs?
Yes, to within 66% probability.
I imagine that at least a few of the graphs would look pretty similar. So I would say “probably not.”
Anyway, since I have been answering your questions, maybe you could try to answer mine.
Given your apparent statement that the Esper 2002 temperature reconstruction is not intended to represent a global average, would agree that it makes no sense at all to splice in the instrumental global temperature average to the Esper 2002 graph?
Given your apparent statement that it’s worthless to ponder over whether the Esper 2002 proxy measurement would have a “blade” if it were extended from 1992 to 2007, do you agree that jshore’s earlier implication that the extension would in fact have such a blade is meaningless?
Please show me a cite that says there is at least a 66% probability that there has been relatively little temperature variation between 1000 and the early 20th century (maximum of 0.5 degrees C from between highest anomaly and lowest anomaly), and I will study it. Thank you.
Why yes, yes I would. Point in fact, I said so many times.
For instance:
He said that Esper’s graph does appear to follow the instrumental average global temperature up to the point at which it cuts off, so it’s reasonable to assume that it would continue to do so.
I would say that basing your results on that would be foolhardy for making a multi-billion dollar decision, but still a reasonable path to follow if you are simply responding to some fool who is trying to prove that because one local area graph doesn’t appear to have an upturn, that this proves anything. Of the options, it is more likely that Esper’s would follow the instrumental temperature from a basic eyeballing. “More likely” might only be 60%, but it is still the winner.
http://boards.straightdope.com/sdmb/showpost.php?p=9164082&postcount=8
Question Three:
If I was to ask 100 people to graph their view of a marble falling through a vacuum, but where each of them had to view the fall through a different distorted glass, what would you say would be the best way to determine the actual graph from the 100 different results? (Obviously, this would be without knowing the basic physics math to simply simulate the occurance.)
Obviously the best choice is to use the graph that best supports your political biases, ignoring the vast majority of the others that point to something very different.
I thought I’d give Brazil a break and field this one for him.
Then I have no idea what your point was in Post #94.
No, he said this:
Another possible reason it doesn’t have a “blade” is that the proxy isn’t closely measuring the instrumental temperature. After all, if you cut off Mann’s hockey stick at 1992 instead of 1998, it would still be a hockey stick. It’s also possible that the instrumental records are wrong. But that’s not the reason jshore is claiming. He is claiming that the reason there is no blade is because the proxy is cut off at 1992. I am asking for a cite.
Can you quote the language you are relying on? I don’t see it. Thanks.
I would probably ignore the 100 results completely and just derive the path of the marble from physical principles.
Why is that obvious? Anyone who has taken physics in high school or college can look up the necessary formulas in a textbook.
But assuming you are trying to graph the movement of a marble based on peoples’ observations, and there’s no a priori knowledge of the marble’s movements, I don’t see any best way (or any way) to do the problem.
Any other questions?
McIntyre linked to the original report. How is that withholding information?
…
Some of the many problems with Mann’s “adjustment” for the bristlecone problem are discussed here.
Hmmm. The NAS report says:
Now, they didn’t say "the data is neither perfect nor totally discredited, it’s somewhere in between. The didn’t say “use the data but be careful with it”. They didn’t say “the bristlecones are not discredited”, as you say. They said the bristlecone data should be avoided. Sounds pretty binary to me, there’s “should be avoided” and “should not be avoided”, and in this case, they should be avoided.
I find it incredible that you, and Michael Mann, are still arguing that the bristlecone data is somehow valid. The NAS findings regarding bristlecones are supported (and preceded) by other scientists. The NAS said don’t use it. Graybill and Idso said don’t use it. Rob Wilson said don’t use it. Heck, Biondi et al (including MBH co-author Hughes) said (emphasis mine):
Why is that so hard for you to understand?
Yes, I have commented before on the convenient way that time works in your world, you seem to have time for everything except things that might disagree with your ideas … in particular, the idea that the synthetic data “would not have actually passed [Mann’s] significance tests” is laughable, given Mann’s refusal to release the results of his significance testing of his own results, so I can see how you might not have time to investigate that aspect of the discussion …
Since Mann did not give an “assessment of the uncertainties regarding the results” resulting from including data which the NAS says do not use, I fear I don’t have any idea what this last point means. A simple check would have been to redo his calculations without the bristlecones, to incorporate that uncertainty … but ooops, we can’t look at that result, because Mann either didn’t do it, or didn’t report it if he did do it.
The Wegman report said:
Seems pretty clear to me … according to one of the nation’s premier statisticians, Mann’s analysis doesn’t support his claims.
Finally, I note that you have not challenged any of the facts offered in the cited thread. The NAS committee and the Wegman Reports did not find a single error in McIntyre’s analysis of the Mann paper. Not one. Take that as your starting point.
You say above:
The problem is that the Mann method mines for hockeysticks, as you agree. Combine that with a single piece of bad data, like the stripbark pines, and presto! Out pops a hockeystick. Since the method mines for hockeysticks, and the dataset supported that mining, I’d like some evidence that it did not happen in this case as you claim. The fact that other methods can replicate the hockeystick only means that those methods are depending on the bad data as well, it does not mean that Mann’s method did not mine for a hockeystick in this case.
I close by repeating Wegman’s statement:
If you want to argue with Wegman about how Mann’s analysis is not statistical garbage, be my guest … until then, I’ll take his statement as the final word. Mann’s analysis doesn’t support his results.
w.
Because otherwise your answer to the question would have no bearing on the topic at hand. (I’m not asking these questions just for heck of it.)
Please answer the question, making the assumption that the movement of the marble cannot be determined except through observation.