Wednesday, February 14, 2007

Lies, Damn Lies, and (climate) Statistics

Arnold Kling has this piece on TCS today. My favorite line is, "Climate data looks to me suspiciously like macroeconomic data. The true information content probably is not sufficient to produce a reliable model for forecasting." Having worked, both in school and professionally, with large data sets and economic models, I have always felt that climatologists either exaggerate or misunderstand the predictive value of their models. One of the concepts that was drilled into me in school is that even the best statistical models are only explanative, not predictive. In plain English, that means, they help describe how different variables were related in the past, but they have very limited ability to tell us how complex systems will interact in the future.

However, like Mayor Giuliani, I think that some of the ideas being kicked around to prevent the spectre of global warming are good for other reasons.

That's why I was glad yesterday when I noticed this story on Instapundit. Somehow it seemed like a strangely familiar idea.

Finally, as always, it is nice to see new additions to Professor Mankiw's Pigou Club.

No comments: