Another hockey stick found in the southern hemisphere

Oh, rubbish!

Mann applied the very same PC selection algorithm to short-centered and full-centered SVD outputs and got almost identical results. The number of PC's retained is anything but arbitrary. The number of PC's that Mann retained was determined by a specific mathematical algorithm -- for both the short-centered and proper full-centered cases.

Furthermore, it is quite easy to determine when you have retained the appropriate number of PC's. You simply repeat the computations, each time adding in another PC. When additional PC's have little impact on the results, then you know that you have retained a sufficient number.

It's quite like a Fourier-series representation of a particular waveform. How many Fourier coefficients do you need to reproduce the waveform reasonably closely? Simple -- just add in Fourier coefficients 1 by 1 until new coefficients no longer change the waveform representation significantly.

The bottom line is, if an additional PC changes your results significantly, then that's a slam-dunk indication that you didn't originally retain a sufficient number of PC's.

 

The answer to that question is simple. If the 4th PC significantly changes the shape of your reconstruction then there is a problem with your method because the 4th PC shouldn't significantly change the shape of the reconstruction. The problem with the Mannian method when dealing with properly centered PCs is the reweighing of the series based on their temperature correlation instead of their eigenvalue. This is selection based in the dependent variable aka double dipping which will inherently create spurious trends that are consistent with the selection criteria.

I of course went over this thoroughly in my first post as I repeated said it is a product of over selecting PCs and then reweighing according to temperature and not eigenvalue which undoes the PC analysis. You chose to ignore the and then part and only address the former like the true Mannian warmmonger.. Dr. Mann would be proud of you.

 

I do not doubt that they will be able to modify their method after the fact to create yet another hockey stick. They can simply choose to throw out the detrending in which case this becomes yet another correlative model. However, as was even confirmed by the NAS panel choosing a method to reach the desired conclusion has no scientific validity.

 

Er, no. Not in a legal or moral sense. The university owns the data, in both a legal and moral sense. You can't simply FOIA everything a university program does because it touched a government grant. I mean, you can try, but the university is under no legal or ethical obligation to comply.

You can certainly claim data loaned to you by others on the condition it not be shared is proprietary. FOIA's do not magically dissolve such agreements. Again, just because an FOIA asks for it does not mean one is legally obligated to supply it.

Ah, the sweet smell of unsupported conspiracy theories. After your pals got caught lying so shamelessly in the denialist Climategate faux-scandal, honest people have renounced them. All that's left now are the ones too emotionally invested in the issue to admit how hard they fell hard for the denialist lies.

It's not going to get any easier for you. You've already had to abandon "There's no warming" and "well, there is warming, but we didn't cause it." What's your next line of defense? Are you going to make a stand at "The warming won't be that bad", or retreat further to "It will cost too much, so we can't do anything."? You must know you're going to have to retreat again eventually, so you should plan for it now.

 

I actually want to address this since I do a lot of work with Fourier series. This is without a doubt one of the most asinine arguments I have ever read. If the nth coefficient changes the wave from so significantly as to change its very nature that is not evidence that the nth coefficient was that significant and that the final waveform is correct but that you screwed up. The nth coefficient should never ever change the shape of the waveform it should cause very slight characteristics of the waveform. If I had a clear sawtooth at the n-1 coefficient and then had a square wave at the nth coefficient I made an error somewhere because the nth coefficient should never have such an effect on the final waveform. If the nth coefficient does have such an effect that isn't evidence of accuracy but of your own incompetence.

 

Nope. You retain whatever number of PCs are needed to differentiate signal from noise. MBH98 used the standard criterion, Preisendorfer's Rule N, and retained 2 PCs. McIntyre re-centered the PCs but still retained just two, even though according to Rule N, after re-centering he should have retained 5 PCs. McIntyre's response to criticism on this point was not "gee, I was wrong, let me re-do it." That would have been the honest thing to do (and that's exactly what MBH did in response to McIntyre's criticism). But no, McIntyre's response was to trash the whole idea of using Preisendorfer's Rule N to begin with, so he could pretend he didn't make a mistake.

 

Well that doesn't really address my point of the PC 4 being responsible for the shape of the graph but let me address it anyways.

Oh please as if the warmmongers haven't' raped Preisendorfer enough.

1st. There is no evidence that Mann used Rule N to select his proxies. He claimed he did 7 years after his original paper and 2 years after MM03 was published but it has been impossible to replicate the MBH98 selection using this rule. There is evidence that Mann reweighted the series before selection and speculatively could have applied Rule N then but that would totally void PC analysis and rule N all together.

2nd. M&M never threw Rule N to the trash. They simply read Preisendorfer. Preisendorfer never once said that Rule N alone was to be used to select PCs but just a necessary step.

"The null hypothesis of a dominant variance selection rule [such as Rule N] says that Z is generated by a random process of some specified form, for example a random process that generates equal eigenvalues of the associated scatter [covariance] matrix S…

One may only view the rejection of a null hypothesis as an attention getter, a ringing bell, that says: you may have a non-random process generating your data set Z. The rejection is a signal to look deeper, to test further."

Even if the Graybill series is significant under Rule N that doesn't mean it should make it into the final analysis. Rule N only sets the stage for the scientist or statistician to look further into the series. Rule N allows us to limit analysis to potentially significant series. It doesn't mean that the series is significant and should be included in the reconstruction. After rule N is applied and the number of series limited Preisendorfer calls for one look for "physical meaning or for clues to the type of physical process underlying the data set Z."

M&M did this. The Graybill series was a known CO2 proxy and not a valid temperature proxy so they rejected it. Mann should have done so too. At one point Mann did do so. His censored folder contained runs of his algorithm that didn't have the Graybill series. Guess what no hockey stick. At some time while developing the hockey stick Dr. Mann knew full well that the Graybill series wasn't a valid temperature proxy. Perhaps he was actually trying to correctly apply Preisendorfer.

To summarize M&M never rejected rule N. Rule N is not how we select significant series. Rule N limits the possible significant series down to a manageable number so the one doing the analysis can then look into the remaining series and check for significance.

Preisendorfer has been dead for many years. Out of respect could you warmmongers quit raping him.