Fun with curve-fitting

These are the great results you can get if you look backwards at brief intervals of data and don't check your results by hypothesizing a causal mechanism, making a prediction, and finding out whether your curves match into the future.  For instance, there is an uncanny ten-year correlation between the number of letters in the winning word in the Scripps National Spelling Bee and the number of people killed by venomous spiders.

2 comments:

E Hines said...

Along those lines is this claim, made by a female student, that my wife heard in a graduate (!) statistics class: "Whether I have a baby or not is independent of whether my mother has a baby or not."

Eric Hines

Grim said...

That’s a great example. True today, true for all possible tomorrows; and true yesterday, and for a great many yesterdays that might be in your sample! And yet...