RE: reliability
Permalink
Well, I'm not the crazy one :-) I agree that we could have played up your point a bit more in the article, but we decided to keep the math simple as possible. Footnotes (1) and (2) describe the point you're making here which is that 10 is way too small of a sample to obtain proper reliability statistics. Ideally, we'd have thousands of samples. Even the reliability figures that we show you which are based on 30 years of model hindcasts are not long enough and have some degree of error associated with the estimate. Thanks for weighing in and reinforcing this point-- complete with a great joke!