I ran across a nice post on the excellent Open Mind blog that shows with absolute clarity why the "cooling trend" of the last decade or so has no relevance to global warming. I've posted about this before, using a similar technique, but Open Mind goes into considerably more detail.
What's nice about these analyses—Open Mind's and, humbly, my own—is that the demonstrations have nothing specifically to do with global warming. The data are not climate data, and it doesn't matter what you think about climate change. They're just common-sense math. They can't be obfuscated with charges of sensor data inaccuracy or urban heat island effects or global conspiracies of grant-happy scientists or any of that. They are what they are.
Here the Cliff's Notes version of the Open Mind post.
Open Mind's author, Tamino, programs a set of data points to have a small upward trend (simulating global warming) and then superimposes on that a bit of "noise" (random upward and downward deviations, simulating weather). If you look at the whole graph, you can see the trend clearly, despite the noise:
Then he steps into the role of Global Warming Critic and takes a subset of the data, starting with "1998" (see the blog for why Tamino labels this data point as "1998"):
Presto! An instant decade-long cooling trend! Global warming is a hoax!
Well, of course not. The long-term upward trend can't possibly be wrong, because it's built in. It's literally programmed into the model. It's as real as it gets.
What does this show? It shows very, very clearly that noise can easily hide a trend if you choose the right time span. In the case of weather "noise", it takes longer than a decade to average out and give you a true picture of the climate trend.
So, the next time your friendly neighborhood GW critic trots out "It's actually been getting cooler since 1998", you have even better information on your side. Nice job, Open Mind.
P.S.: For yet another take on the same concept, Andy Revkin of the New York Times has a good post on his Dot Earth blog.