I don't think that's cherry picked; it looks like an outright fabrication to me. What are the odds that a line of regression on a graph like that would come out to exactly +0.00C?
Out of a sufficiently noisy continuous dataset, you can actually do that quite easily simply by moving the start position until you hit exactly +0.00C. Which is what I suspect he did. If you move it forwards a few months, you get negatives until you leave that one anomalous spike out, and if you move it backwards, you start to get positives very quickly.
Tuna-Fish|11 years ago