## Section 4.6 Responses

4. It is alleged that there has been an improper bias in selecting and adjusting data so as to favour the anthropogenic global warming hypothesis and details of sites and the data adjustments have not been made adequately available It is alleged that instrumental data has been selected preferentially to include data from warmer, urban in contrast to rural sites; that the rationale for the choice of high/low latitude sites is poor; and that the processes by which data has been corrected, accepted and rejected are complex and unclear.

QUESTIONS TO ADDRESS

6. What means do you use to test the coherence of the datasets?

February 21, 2010 at 3:05 pm |

Matbe here for a technical discussion of the statistics commonly used for detrending, autocorrelation

http://listserv.arizona.edu/cgi-bin/wa?A2=ind1001&L=itrdbfor&T=0&P=5570

Assuming you want to relate these ring widths to something else, like

climate data, you have to somehow remove the autocorrelation structure from

each core (or tree), by removing the age-related trend. You can use a

linear filter, a function that relates ring width to cambial age (e.g.

negative exponential or linear equation), a cubic spline, an autogressive

model, or something like a first difference, depending on the

autocorrelation structure of the ring widths The goal is to obtain as

serially uncorrelated a series as possible, for each series. You can then

calculate any statistic you want from the resulting data, with n = 1000 in

your example. You can also detrend the mean of the 10 series instead, and n

would then = 100. Keep in mind though, that the relationship with the,

e.g. climate, data will very much depend on how you detrend the ring data.

The other way is to simply calculate the lag 1 autoregressive coefficient

(r) and plug it into an equation that computes the effective number of

degrees of freedom, which is of the form Neffective = (1-r)/(1+r). That

assumes that the autocorrelation structure is dominated by a lag 1 effect.

Your standard error calculation is then based on Neffective, not the

original n.