Please keep up, and show out, this work; struck anvil with law.

============================== ]]>

“How does your analysis fit with the authors claims of confidence? Is there any trends in the data that are better than noise?”

Ron, I would have to combine the proxy data altogether in bins in the weighted manner I used for individual proxies (using anomalies) and from that determine whether a composite trend in the manner of McGregor was statistically significant. Since I think the long term trends are an artifact even statistical significance would have little meaning for me. I count 39 out of 57 proxies having long term straight line trends that are significant using p.value less than 0.05. Of those 37 proxies 29 are negative by my count. I suspect that one might obtain a significant negative trend for the composite although since the individual proxies have different time periods that is an iffy guess.

I have started a serious and detailed analyses of the temperature series from the observed series and CMIP5 climate models using singular spectrum analysis and diagnostics, but I might go back and look at the McGregor binned composite. These calculations are never as easy as they seem prior to doing them.

]]>Ken, I understand that you are testing the data in different statistical models and for things like auto-correlation. I take your replies to my questions regarding any oscillation signals as there are none. Can you explain your sentence in more detail: “These analyses results add further evidence that there exists an unrealistically good fits of the proxy data over long periods of time to linear trends of differing slope values.”

How does your analysis fit with the authors claims of confidence? Is there any trends in the data that are better than noise? I studied the plots on your Excel sheet for a some time last night and I withdraw my first impression that there are any oscillations that have any pattern or correlation. I was particularly impressed by the lack of correlation of study pairs that we done in the same location. With Bonnet’s dinocyst and Spielhagen’s planctin foramin studies side by side in the same location in the arctic there was an opportunity to validate two different proxy types. They look to have zero correlation. And they do not match what should be general warming in the arctic, with Bonnet’s 2K low in 1927. Spielhagen spikes up at the end but at 1800, well before modern AGW.

In most studies it is hard to gauge precision with such space sampling. Black, however, with over a 100 samples reveals the pretty mediocre precision, at least with Mg/Ca. Lea does a study in the same spot but they did not think to overlap their time series for a concurrent validation, apparently. Not that I found any correlation in studies done in the same location or approximate location by the same researcher. I looked at pairs from Saenger, Pahnke, Lund, Nieto/Moreno and Richey.

Pahnke did side-by-side 110-yr and 160-yr studies, showing an attempt to get high resolution in alkenones, and though they both upward trends, one peaked in 1925, the other in 1999 but after being at a plateau since 1925. Both were marked in 0.01C precision, one having a 0.15C 100-yr rise and the other a 0.50C 100-yr rise. Again, in locations only separated by 3.5 minutes of latitude and 1.8 minutes of longitude.

]]>Ron, I thank you for your thought provoking replies. Without it I would be pitching a shutout on responses – not that it has not happened before.

]]>Ken, it sounds like you have some very interesting findings. I think many are eager for one of the twelve people on the planet that can translate the implications to stop by.

]]>I have added, in the link below to an Excel file previously linked, the QQ plots of the arfima residuals of the 57 binned McGregor regression residuals showing that the arfima residuals have normal distributions with the exception of 7 proxy arfima residual series which fail the shapiro test for normality. The results of shapiro test have also been added to the Excel file linked below. In most cases of the failure of shapiro test for normality can be seen in the QQ plots to be due to 1 or 2 outliers. These analyses results add further evidence that there exists an unrealistically good fits of the proxy data over long periods of time to linear trends of differing slope values. These good fits make the fits suspect as an artifact of the sedimentation process and/or how the sampling was performed for the 57 McGregor proxies.

I have been disappointed by the lack of response to my analyses shown on this thread.

]]>I have added 2 worksheets to the original dropbox Excel file and linked it below. The 2nd worksheet shows the result of a spectral analysis on the 57 McGregor linear regression residuals and shows no cyclical tendencies as noted by comparison to the results obtained from ARMA(0,0), ARMA(1,0) and ARMA(2,0) simulations.

The arfima analysis shown in the 3rd worksheet shows that 10 of the 57 McGregor linear regression residual series give a somewhat improved Box.test score when a fractional d is included in the ARMA model.

]]>Okay, I see oscillations. On more study they are not consistent enough to be solar caused. But most of these plots in your dropbox have some wiggle that has some signal. If it is not caused by systematic error in the analysis it is easily the most important information in the study. First, if there is truly a natural oscillation this would be a new one, not tied to ENSO and too low frequency to be AMO/PDO but supporting the concept of a multiple harmonic based climate. Second, a local climate change in the oceans on a 200-yr +/-100yrs means local land climate is not only possible but expected and cyclical. Third, if oscillations are true then they provide an intrinsic validation of the proxy studies precision. Fourth, if the proxies are valid the HADSST is not.

Of course if you’re right, and there’s nothing there, never mind.

]]>Ron it is probably just you. Recall that 30 of the proxies had random normal distributions for the regression residuals. The remainder were well fitted with ARMA models. ARMA models can handle some cyclical properties but a regular cycle would require a high order ar to obtain a fit.

I can look at a spectral analysis and see what results. I also am reminded that I could see how a long term persistence model fits with ARFIMA.

I think you may have qualified as a prospect for a temperature reconstructionists as you appear to see patterns in proxy series before validating a true proxy response.

]]>Ken, is it just me or is there a 200-year oscillation showing in most of these proxies? If I am not seeing things can you quantify such a signal with Fourier transform analysis? If it is real the solar guys would flip. Some claim a 207-year solar cycle exists.

]]>