I’m in your data, busting your analysis
One of my advisors in college wisely told me, “More math will never hurt you.” Even though they made me crazy, nearly ruined my GPA, and seemed to be taught by different versions of the same person, I’m glad I took all those stats classes. While I don’t run chi-squared tests on a daily basis, I do have a pretty finely-honed bullsh*t detector. Take, for example, data contained within the Mothman Prophecies:
I was in the mood for a little light reading, so I picked up this 1975 gem on unexplained phenomena. Perhaps it’s all the Penn & Teller I’ve watched, but whenever someone in a “woo-woo” field cracks out data, I just have to look under the hood.
I do give Keel credit for actually taking a crack at being scientific and keeping data. He seems reasonably objective and who knows how I’d think if faced with the Mothman itself. That said, relying on straight-up percentages can get you in trouble.
On page 144, he displays a table of distribution of UFO reports in 1950 and concludes, based on percentage of total observations, that UFOs are more likely to be reported on Wednesday than any other day of the week, since 18.8% of observations reported were on Wednesdays.
It’s been at least eight years since I’ve run any kind of statistical test, but I just couldn’t help myself. Wikipedia made un-sticking the gears a relatively painless procedure, and I was soon using Excel to run a chi-squared test (which I think was the right one to use).
Result: I reject Keel’s hypothesis and find that a UFO is no more likely to be reported on Wednesday than any other day of the week.
And that is why I only went on four dates in college.







