Tuesday, September 05, 2006

You could fit a cow with that

In the recent gale force back and forth about hurricanes at Volkh, something important was been lost, something that the Bill Grays of the world would soon see stay lost...

To more or less (mostly more) quote myself, everyone goes on as if there is only data. Analysis with only data is curve fitting. snark: Analysis with only models is string theory /snark.

We have the assumption of an Atlantic multi-decadal cycle in hurricane intensity and frequency, but THIS IS CURVE FITTING, and, from the data, how many cycles have been observed? Think about it, how do you fit a cycle if you have not accurately observed many ups and downs. Worse, how does this apply to the Pacific. To give you an idea of the mess this curve fitting gets you into we have the words of Len Pietrafesa, hurricane expert at North Carolina State University:

"There are five fundamental modes of oscillation -- three to five years, 10 to 12 years, about 20 years, about 35 years and 45 to 60 years. The (oscillation) modes superimpose, and we get what we get as a function of the year we are in."
You could fit a cow with that number of Fourier components.

No comments: