Log in

No account? Create an account
Thoughts Online Magazine
Collected Articles on Culture & Politics
Statistical Phantoms 
5th-Jan-2007 05:37 am
Once again, I got an email from someone asking what a "statistical phantom" was. Evidently, whenever I give this talk, someone is taking notes, and they talk at the local coffee shop with someone else, who talks....

Well, here goes. A statistical phantom is typically a correlation, average, or other measure which has been applied without considering whether or not it is appropriate. For example, bikini sales and ice cream sales seem highly correlated: but the reason has neither do do with bikinis nor cones, but to do with summer, whether we are cooling things down or heating them up. People are fooled by statistical phantoms all the time, and there are researchers who get grant money because they present correlations and ask for funding to investigate the cause, after sending press releases to their local, gullible reporter. You might, if properly educated, have heard the phrase, "Correlation does not imply causation." Correlation may not: but newspapers, and poor researchers, get paid to.

An example? The Times of London conveniently provides one. The promoter of the study has his quasi-scientific explanation at the ready: "Where and when we are born does have an influence on our futures,” says Dr George Davis, the physician and scientist who led the studies. “This is not astrology, this is basic science. The sun has been around for a few billion years,” says Dr Davis. “It would be surprising if life did not march to its beat; 98 per cent of the time it is benevolent, but 1 or 2 per cent of the time it is destructive.” The rest of the article is a collection of spurious correlations. My favorite is this:

Dental disease
A Harvard University study showed that dental disease increased the farther you are from the Equator, and the coast. The researchers say sunshine, temperature, rainfall and humidity seem to be factors.

Just as a guess, where are the priciest communities in the US? Oh, the coasts, you say? And so they might be likely to attract good dentists, eager to have paying clients? You know, subtropical rain forests never seemed like a good location for a dental office.... A spurious correlation, a statistical phantom. And yet, as noted, Harvard University has people who got grants to study it.

Is Harvard University a haven for cranks and crackpots of all kind? No more, I suppose, than the other universities. We have to keep them someplace: otherwise they would be out wandering the streets, and might get run over by a bus, snarling traffic. Think of it as a partial solution to the homeless problem.
This page was loaded Jan 18th 2019, 1:58 pm GMT.