CERN: Climate Models will need to be revised

[Michael Hardner]

It is not an allegation. It is a fact that is clearly stated in the paper.

I am not sure there is any point to continue if you refuse to acknowledge that certain points are facts and are not open to interpretation.

In Mann's paper or Tiljander's ?

[Michael Hardner]

His claim now is nothing but an after the fact excuse to cover up his screw up.

How could following a published process be a "screw up" ? Unless they made some other mistake along the way.

If two sets of data result in two different coefficients, then that's what the process came up with, it's not a screw up.

[TimG]

In Mann's paper or Tiljander's ?

Tiljander paper makes it very clear that the data the last 100 years is contaminated. This is a fact. Not an allegation.

[waldo]

If Mann really believed that Tiljander got it wrong he would have had to reverse the signs of both.

In any case, Mann has no business changing the interpretation of the proxy authors.

His claim now is nothing but an after the fact excuse to cover up his screw up.

nice leap... if you can actually read, that was my conjecture offered on possibly some influences on why a two-sided test was deployed.

what's this - you want to hold fast to a Peer-Review interpretation from the proxy author... I will most certainly oblige you, in kind. I'm always quite eager to replay this most insightful directing comment from the proxy author's paper, in her own words:

...the proxy author, Tiljander,

from your declared authoritative source paper

, summarily states:

However, it is difficult to make climatic interpretations at the annual time scale, but short-term changes (averaged over a few years) could be estimated.

yes...

from your declared authoritative source

:

difficult... but could be estimated

; i.e.; difficult... but not impossible.

[TimG]

How could following a published process be a "screw up" ?

The data should not have been included in his reconstruction. If Mann was a honest scientist he would have simply acknowledged this, re-did his reconstruction without the proxies and adjust his conclusions.

The trouble is Mann is not an honest scientist so he misdirects, obfuscates and lies in order to avoid acknowledging the fact that Tiljander does not belong in his reconstruction.

[waldo]

In Mann's paper or Tiljander's?

both - Mann et al most certainly acknowledged a possible concern... and their methodology proceeded to handle the data, to screen it, to validate it, to calibrate it and to sensitivity test it. In the supplementary information document I provided the link to (a short number of posts back) you will see specific reference to a section titled, 'Potential Data Quality Problems' (a subset within the greater 'Sensitivity Analyses (NH Temperatures)' division).

[waldo]

The trouble is Mann is not an honest scientist

so says you as you parrot your hero McIntyre (the darling of the Heartland Institute)... formally call out said claimed dishonesty. Formally issue a challenge to the Mann et al paper. Is there a problem?

[Michael Hardner]

both - Mann et al most certainly acknowledged a possible concern...

I thought I had read that... yes...

and their methodology proceeded to handle the data, to screen it, to validate it, to calibrate it and to sensitivity test it. In the supplementary information document I provided the link to (a short number of posts back) you will see specific reference to a section titled, 'Potential Data Quality Problems' (a subset within the greater 'Sensitivity Analyses (NH Temperatures)' division).

Yes, I read it.

[Michael Hardner]

If Mann was a honest scientist he would have simply acknowledged this, re-did his reconstruction without the proxies and adjust his conclusions.

If he was a dishonest scientist, then he wouldn't have acknowledged the problem in his notes on the paper.

I feel like I'm getting close to this: The issue seems to be that two different data series gave two differently signed coefficients, both using a valid regression analysis (or similar process).

So either one or other other used their process incorrectly, or the process itself that either one used is wrong, or (what it seems like to me) this is just a curiosity that comes from the different data. Which one is right ? These are, in the end, two models.

Maybe there are cofactors that are causing the sign to flip, such as rainfall as has been pointed to ?

[TimG]

If he was a dishonest scientist, then he wouldn't have acknowledged the problem in his notes on the paper.

His notes show that he was either too stupid to understand how the contamination makes it impossible to use the data with his correlation algorithm or that he knew that and he hoped to fool people with a little bit of hand waving. I tend to assume that Mann is not stupid therefore the only rational conclusion is he is dishonest.

I feel like I'm getting close to this: The issue seems to be that two different data series gave two differently signed coefficients, both using a valid regression analysis (or similar process).

You are getting lost in the weeds. The issue is screening with correlation does not work if the data is extremely contaminated in the period used for the screening. The models that are added on afterward are irrelevant. Bonam made basically the same point here:

http://www.mapleleafweb.com/forums//index.php?showtopic=19451&view=findpost&p=774555

Edited March 19, 2012 by TimG

[waldo]

If he was a dishonest scientist, then he wouldn't have acknowledged the problem in his notes on the paper.

obviously... the Mann et al authors acknowledge the potential problem and quote directly from the proxy author - hardly the under-handed, devious, dishonest approach TimG blathers on about. Quoted proxy author words such as, "human impacts have distorted the natural signal to varying extents" and "a demanding task to calibrate the physical varve data". Yet, again, as I highlighted, and will continue to highlight, the proxy author also states, "However, it is difficult to make climatic interpretations at the annual time scale, but short-term changes (averaged over a few years) could be estimated." Equally, the TimG false narrative continually targets the single person, the single scientist... purposely choosing to ignore that the Mann et al paper has 7 authors, including 4 who have significant and recognized careers, awards, associations and publication histories with hundreds of published papers. I expect TimG will begin to forcefully trot out his oft repeated conspiracy themes... that "all these authors" are conspiring together!

I feel like I'm getting close to this: The issue seems to be that two different data series gave two differently signed coefficients, both using a valid regression analysis (or similar process).

So either one or other other used their process incorrectly, or the process itself that either one used is wrong, or (what it seems like to me) this is just a curiosity that comes from the different data. Which one is right ? These are, in the end, two models.

Maybe there are cofactors that are causing the sign to flip, such as rainfall as has been pointed to ?

I infer you're drawing comparison between CPS & EIV... I would suggest you're not following the distinction as to the nature of datasets processed by each, compared estimate similarities and differences... and what differences mean. As Dr. Mann responds in regards to a question concerning CPS vs. EIV result differences:

The EIV and CPS approaches, as described and shown in the manuscript, produce results that are (i) remarkably similar back to AD 1500 or so, (ii) certainly consistent within estimated uncertainties back to AD 1000 and which finally (iii) do differ outside the uncertainties prior to that. Where the estimates differ, it means the data are sparse enough that the differing assumptions underlying the two methods really do matter, and not surprisingly the answer does depend on those assumptions. This is all discussed in the paper, and even more extensively addressed with parallel 'pseudoproxy' experiments in the supplementary information which test the sensitivity of the two different methods to increasingly sparse data networks

[Michael Hardner]

I infer you're drawing comparison between CPS & EIV...

Thought I replied already... Mann and Taljinder also used different methodologies I take it...

Which one used which ? I can research these independently.

[TimG]

Thought I replied already... Mann and Taljinder also used different methodologies I take it...

You comparing apples and oranges. Tiljander collected thickness and xray data from sediments. The only claim that Tiljander made was that temperatures changed over time - Tiljander did not attempt to define a numerical relationship between the proxy and degC - Tiljander only specified the sign of this relationship.

Mann needed to calibrate all the proxies together so they could used together. To do this he had to define a numerical relationship between the proxy and degC before he could use his algorithms. This is where Mann failed. It is simply not possible to determine any numerical relationship between the proxy and temperature because of the data contamination.

IOW - you are wasting your time looking at what CPS/EIV or any other later processing Mann did because it is irrelevant to the point I am making. What matters is the screening algorithm which is really nothing but a correlation calculation and that calculation incorrectly selected the proxies.

Edited March 22, 2012 by TimG

[Michael Hardner]

Mann needed to calibrate all the proxies together so they could used together. To do this he had to define a numerical relationship between the proxy and degC before he could use his algorithms. This is where Mann failed. It is simply not possible to determine any numerical relationship between the proxy and temperature because of the data contamination.

IOW - you are wasting your time looking at what CPS/EIV or any other later processing Mann did because it is irrelevant to the point I am making. What matters is the screening algorithm which is really nothing but a correlation calculation.

And what matters about that is the contamination ? Mann acknowledged it, so there must be some uncertainty as to how much it matters ?

[TimG]

And what matters about that is the contamination ? Mann acknowledged it, so there must be some uncertainty as to how much it matters ?

There is no uncertainty. It matters. The contamination is so large that the apparent relationship between the proxy and temps changes because of the contamination!

This is something that you should understand automatically if you understand correlation and its limitations. This is something that Bonam realized as soon as he looked at the paper so you are not simply taking my word for it.

I suspect that your roadblock is you want to believe that there was a good reason for what Mann did. There is not. If he had actually took the contamination into account he would have realized that he could not use the data. But he wanted to use the data because it was essential to his 'non tree ring' reconstruction. So he decided to use it anyways and hoped he could bluster and BS. Unfortunately, he can get away with bluster and BS because there are way too many people like you who put scientists on pedestals and treat them as infalliable gods.

Edited March 22, 2012 by TimG

[Michael Hardner]

I suspect that your roadblock is you want to believe that there was a good reason for what Mann did. There is not.

Maybe so, but the fact that he acknowledged the issues with the data make it clear that it's not a dumb mistake. And the fact that the correlation changed, while curious, is not evidence that there was a mistake made.

You can put this on me having blind faith in scientists if you like, but the fact that he looked at this and acknowledged it means it was not deception, nor was it doltish blunder.

Now... I'm looking to understand those two methods. If I can understand the confidence levels, and correlations that were developed it might help me to understand further.

[TimG]

Maybe so, but the fact that he acknowledged the issues with the data make it clear that it's not a dumb mistake.

I was trying to Mann the benefit of the doubt. I think this is really a case of deliberate deception on the part of Mann who really needed to include this particular proxy in his paper but I can't prove it. I can prove it is at least a 'dumb mistake' but the proof requires someone who understands what correlation is.

And the fact that the correlation changed, while curious, is not evidence that there was a mistake made.

Yes it is. If you believe otherwise then please explain exactly how correlation could possibly work with contamination so large the sign changes. I don't believe you can do it.

the fact that he looked at this and acknowledged it means it was not deception, nor was it doltish blunder.

The fact that he looked at means is it most likely deliberate deception. Edited March 22, 2012 by TimG

[Michael Hardner]

I was trying to Mann the benefit of the doubt. I think this is really a case of deliberate deception on the part of Mann who really needed to include this particular proxy in his paper but I can't prove it. I can prove it is at least a 'dumb mistake' but the proof requires someone who understands what correlation is.

If it was a deliberate deception, he could have tried better to bury how this was done, and not called attention to the data issues in his notes.

Yes it is. If you believe otherwise then please explain exactly how correlation could possibly work with contamination so large the sign changes. I don't believe you can do it.

Let's say I was examining the relationship between telephone poles in a neighbourhood and income over different time series. In the 1920s there would be a positive correlation, while in the 2010s it could be negative. Moreover, each correlation could easily be significant.

[TimG]

Let's say I was examining the relationship between telephone poles in a neighbourhood and income over different time series. In the 1920s there would be a positive correlation, while in the 2010s it could be negative. Moreover, each correlation could easily be significant.

You are not answering the same question as Mann. What Mann is doing is using the correlation in the 2010s to determine what the correlation should be in the 1920s. If he did this with your dataset he would end up using the data 'upside down' when he tried to do any analysis on telephone usage vs neighborhood income in the 1920s.

I will reframe my question to suit your example: please explain how someone could use your dataset to predict neighborhood income in 1920s from telephone poles by only using data from 2000 onwards to determine the correlation between poles and income because that is only period where you have neighborhood income data. Would the results of such an analysis be meaningful?

Edited March 22, 2012 by TimG

[TimG]

If it was a deliberate deception, he could have tried better to bury how this was done, and not called attention to the data issues in his notes.

Actually, it is almost Machiavellian. If he said nothing he would have be called on it and he would have no excuse. But mentioning and doing absolutely nothing about it he created doubt and gets legions of people to defend his dishonesty because 'he mentioned it'.

Why can't you see that 'mentioning it' means nothing if he turns around and ignores the problem?

Edited March 22, 2012 by TimG

[Michael Hardner]

I will reframe my question to suit your example: please explain how someone could use your dataset to predict neighborhood income in 1920s from telephone poles by only using data from 2000 onwards to determine the correlation between poles and income because that is only period where you have neighborhood income data. Would the results of such an analysis be meaningful?

No, not in my example.

[Michael Hardner]

Why can't you see that 'mentioning it' means nothing if he turns around and ignores the problem?

That's why I need to understand more about this.

[waldo]

Thought I replied already... Mann and Taljinder also used different methodologies I take it...

Which one used which ? I can research these independently.

And what matters about that is the contamination? Mann acknowledged it, so there must be some uncertainty as to how much it matters ?

Michael... the proxy author, Tijander, did not carry forward to reconstruct a temperature series - at the time of her paper she was only a graduate student... I question whether she had/has the expertise or perhaps interest to pursue large-scale multiproxy temperature reconstructions.

the disingenuous TimG continues to blather nonsensically on. I've now repeatedly suggested you keep your attention directed toward the calibration/weighting - just how significant is the so-called/presumed contamination segment of the overall proxy timeline? Hey now... as I have also repeatedly suggested to you, sensitivity testing is key to determining just how significant the small contamination segment (in the overall) is... or more directly/pointedly, just how significant are the 4 Tiljander proxies to the overall reconstruction? Again, sensitivity testing! All points I highlighted way back in a prior MLW post related to another of TimG's flights of his parroted McIntyre obsession:

the proxy author did not construct a temperature series - none exists. As I'm aware, no chemical analysis of the 4 proxies has ever been done. Certainly, given the proxy authors paper, questions exist over possible contamination of the proxies in most recent years (related to human infrastructure related activity). The Mann08 paper thoroughly acknowledged questions concerning data quality... proxies passed the papers screening processing and were calibrated, accordingly. The only question to arise is whether or not the calibration is on/close across the full range... one aspect of testing the sensitivity of this is to, quite obviously, check the significance of the overall reconstruction with the proxies in compared to their exclusion... this was done within the Mann08 paper, as repeatedly stated to you, over and over. There was no significant affect on the overall reconstruction if the 4 proxies were left in... or removed. Again, no significant affect - no appreciable difference.

The "doesn't matter" context is in terms of the overall reconstruction impact... the 4 proxies in versus the 4 proxies out... what discernible difference can be shown in comparing the in vs. out. As stated, now too, too, many times, Mann2008 - SI - Fig. S8 shows that comparison; i.e., no discernible difference to the reconstruction.

[TimG]

No, not in my example.

Great! Then please explain how someone could use your the Tiljander dataset to predict neighborhood income temperatures in 1920s 1000s from telephone poles varve x-ray density by only using data from 2000 1850 onwards to determine the correlation between poles density and income temperature because that is only period where you have neighborhood income temperature data. Would the results of such an analysis be meaningful? Edited March 23, 2012 by TimG

[waldo]

That's why I need to understand more about this.

your understanding 'more of this' won't be realized by you continuing to coddle to TimG's nonsense.

TimG can continue to bark loudly with his unsubstantiated claims of dishonesty, of fraud, of deceit, of cunning, of duplicity... of the "Machiavellian Mann" - at the end of the day TimG's parroting blather, the following remains firmly entrenched in the WDC/ICSU WDS world archive of Paleoclimatology Climate Reconstructions => Proxy-based reconstructions of hemispheric and global surface temperature variations over the past two millennia

in over 3.5 years nothing other than the likes of TimG's parroting shriek has come forward to formally challenge... what are "they" waiting for... is there a problem?

No formal challenge in 3.5+ years <=> TimG nothingness