Anyone feel like taking a stab at what statistical sleight of hand has been used in constructing this graph?
http://www.openmarket.org/wp-content/uploads/2009/09/christy-observations-vs-hansen-1988-models.bmp
Why not a more modern version of the graphic by Christy? (now without Roy Spencer)
We have two years more to enjoy.
Just for starters, those graphics are just what you get when you compare pears and peaches to measure the state of the dynamical fruitness.
Christy uses his own controversial UAH "observational" dataset, the lower troposphere (LT) temperatures that, as everybody knows, are exactly the same as surface temperatures

, the same way thermometers placed on your forehead, in your ear canal, your armpit or Uranus give exactly the same value, the same temperature you have in your Broca's area, your rightmost toe (the leftmost is a rebel) or around the healthy linseeds your guts are trying to digest to no avail.
Of course, those UAH-LT annual means are validated

with the dataset RSS-LT which is basically the same manure but with a different scent and not surprisingly it gives similar results.
Then, you have to compare the prophecies

by that devil Hansen -who is sleeping now in his coffin because there's daylight out there- in his Senate testimony during 1988:
When you compare both sets of values the first question that emerges is what is the zero value in both sets. The Hansen's zero is 1950-1980 normals (surface temperature). What values did Christy have previous to 1978? His values suspiciously match the GISS surface
modelled temperature dataset at the beginning of his series. Maybe, there's nothing to suspect about because that old model was validated against existing datasets so the model can't differ much from reality around 1980.
The deceiving part of the graphic is the use of "observational" against model: the model never intended to measure the same temperatures that Christy "observes".
About the model, consistent with a time where publications looked like the image containing Hansen's, the Internet was just a nerdy thing with four-eyes using gopher and Mosaic, and a 286 with 2 Mbyte of RAM and 35MHz of clock, and a cache of 128 KByte were the space rockets of the time, not only we can look at it with a forgiving smile, like we'd looked Gutenberg when compared with a laser printer, but we can also see that Hansen may have committed the little sin of choosing individual runs of the model just to avoid showing the lack of intellectual abilities of the senators by using bands and error margins. You can see the saw-like aspect shared by recorded series and individual model runs -likeness that Christy exploits in his favour-. One or two series may contain a volcanic event by 2014, so one may think that Hansen didn't choose runs that matched the average in the end.
There's more to say, but this is a futile exercise that combines old deprecated modelling with controversial datasets in an anecdotal context, all with a few graphical and data-selection tweaks to show what they want to show: damn lies and statistics (besides, the troll has returned)