Re: jae (#57),

There indeed is a measure of skill that seems to have escaped the notice of paleoclimatologists. It is the cross-spectral coherence squared, which shows the fraction of variance in each frequency band that is inter-related between two time series. I’ve already offered Steve Mc to cross-spectrum analyze any pair of proxy and calibration data series that he sends me. A by-product of such analysis is the cross-covariance matrix for optimal reconstruction filters (Kalman-Bucy) along with an explicit measure of their potential power. I’m expanding my offer to analyze any two proxy series (the longer the better) to reveal their respective spectral structures and phase inter-relationships. That should answer the questions about fairly regular oscillations evident in many proxies raised by other commenters.

]]>replace

for n=1:19 %loop over each proxy set (back to year n*100)

by

for n=1:9

etc, for example. Now NH_whole_newgrid_HAD_gbeachyear.mat gbcpsfull(:,4) agrees with archived NH_had.mat gbcps for 1500-1600, so I guess I’m on the right track.

What does gridboxcps.m do ?

* Each gridbox instrumental is filtered with Mann’s smoothing algorithm

* Proxies are standardized (they seem to be Mann-smoothed already in griproxy.m; cutfreq 0.1 for ‘annual’ and 0.05 for ‘decadal’ proxies)

* All proxies in 5×5 grid are averaged

* Within the grid, proxy set is ‘calibrated’ with variance matching to temperature. Smooth first, calibrate then. What would happen vice versa ? I have to read ‘The Foundations of Variance Matching’ again..

(* Data is regridded to 15X15)

* Grid-reconstructions are area-weighted to get the hemispheric means, and variance matched again (!!) (with CRU or HAD as target).

]]>??? Out of memory. Type HELP MEMORY for your options.

Error in ==> gridboxcps at 132 saveregts=reg_ts(:,igood,:);

reg_ts seems to be too big for my computer

reg_ts 1996x2592x19 786392064 double

and seems to be almost full of NaNs. Any volunteers to optimize the code by applying sparse matrices ?

]]>Re: bender (#116),

Mann’s errors and shiftiness are probably embarrassing to have to continually defend. I feel sorry for Schmidt for having chosen such an undependable character as a trusted colleague. RC should throw Mann overboard.

Agree. The game gets tough for Gavin. Quite a work to go through different blogs such as

http://wmbriggs.com/blog/2008/09/06/do-not-smooth-times-series-you-hockey-puck/

and try to defend Mann’s methods. And this is just the beginning, Mann08 requires some time to comprehend. Code with lines

pp=load(‘c:\scozztemann3\newtemp\CRU_NH_reform’);

and

nnn=load(‘/holocene/s1/zuz10/work1/temann/newtemp/HAD_NH_reform’)

makes the task difficult. Specially, I’d like to find

c:\scozztemann3\newtemp\nhhinfxxxhad

( Funny, variance matching is well alive, SI:

The gridded proxy data were then areally weighted, spatially averaged over the target hemisphere, and scaled to have the same mean and decadal standard deviation (we refer to the latter as â€˜â€˜variance matching’) as the target hemispheric (NH or SH) mean temperature series.

)

]]>Mark

]]>