Extrapolating From One Data Point

I had a friend in the consulting business that used to joke that he preferred to only have one data point when he had a point he wanted to make.  "If you only have one data point, you are free to slam a line through it in any direction and at any slope you want.  Once you have two, you are more constrained."

I am reminded of that story reading Trevor Butterworth's fabulous take down of typically bad media "science" scare story, this one on fireproofing materials in mattresses.  He has a lengthy fisking, but concludes:

What CBS produced is an advertorial for ABC Carpets and Homes, more
suited to a shopping channel. By failing to test any of the claims for
a risk against the science, by using a sample of one self-diagnosed
couple, by testing nothing, and not even bothering to interview someone
from the CPSC, let alone an independent toxicologist, the viewer is
left with the message: buy a bed at ABC if you want to be safe.

  • Anonymous

    My former boss used to say if he could define the experiment, he could prove the earth flat.

  • I believe it was Steven Brust who said, "Everyone extrapolates from a single data point. At least, I do."