Roger wrote:

The message
from Roger contains these words:

I might manage to post as early as tomorrow. OTOH I might not. :-)

Or even today. Figures below are for the unsmoothed annual anomaly from
1850 to 2007. I also have the figures for the smoothed curve if anyone
is interested. I used them and the smoothing filter to check for
extraction errors. I think most of the figures are accurate to 0.01 but
I suspect there may be some minor distortion in the graph which might
make matters worse.

The smoothed curve and my smoothing exercise have a reasonable fit with
only 5 years differing by 0.01 or more. (1850 - 0.023,1851 - 0.14, 1955
& 1956 - 0.01 opposite signs), 2007 - 0.014).

Over to you Terry.

A labour of love....

Thank goodness Excel allows one to fill in a data series; I'd have
hated to have typed in all the years from 1850 - 2007 :-(

I've carried out the various analyses that Excel allows: moving
average of chosen span; polynomials of various orders; linear
regression; and logarithmic.

The last two produced silly trendlines and weren't proceded with.

Polynomials of second and third order produced interesting trendlines.

Moving averages of 3 and 10 points were carried out.

Results he

10 point moving average:

http://farm4.static.flickr.com/3179/...4007573e_o.jpg

5 point moving average:

http://farm4.static.flickr.com/3005/...1acc803a_o.jpg

Second-order polynomial:

http://farm4.static.flickr.com/3157/...5f46c52f_o.jpg

Third-order polynomial:

http://farm4.static.flickr.com/3258/...0b5832f2_o.jpg

Points to note:

- only the 5-point moving average picks up a downturn circa 2004 or
so

- the 3rd order polynomial has a slightly higher correlation
coefficient than the 2nd order. The data at the 1850 end of the scale
suggests recover from the Little Ice Age....

Discuss....;-)

[if anyone can do more sophisticated analyses than allowed by Excel,
I'm sure the results would be of interest]