I saw this over at Eurekalert: Charting our Health by the Stars.
This is a press release about new research into Horoscopes’ prediction of health conditions. Or, Not.
“We did this study to prove a larger point â€“ the more we look for patterns, the more likely we are to find them, particularly when we donâ€™t begin with a particular question.â€
Using medical records of millions of Canadians, the researchers basically threw it all in a statistical pot with star signs and stirred it up to find correlations. Indeed, there were definite patterns–apparently Virgos have nasty problems with morning sickness, for example.
Except–correlation is the weakest of statistical relationships. And we’ve all heard “correlation is not causation.” If indeed there were real predictive relationships between horoscope and health, it should work on new population samples.
However, even though each astrological sign appeared to have its own unique disorders, the results were not reproduced when they were tested in a second population. There was no predictive value, and there was no relationship.
Because the author said it so nicely, I’ll quote him:
â€œScientists take pains to make sure their clinical studies are conducted accurately,â€ says Austin, â€œbut sometimes erroneous conclusions will be obtained solely due to chance.â€ Statistical chance means that 5 per cent of the time, scientists will incorrectly conclude that an association exists, when in reality no such association exists in the population that the scientists are studying.
One way to reduce the chances of drawing a wrong conclusion is to try and reproduce unexpected results in further studies.
â€œThere is a danger in basing scientific decisions on the results of one study, particularly if the results were unanticipated or the association was one that we did not initially decide to examine,â€ says Austin. â€œBut when several studies all arrive at similar conclusions, we reduce the risk of arriving at an incorrect outcome.â€
One of the other ways to make your conclusions stronger, statistically and inference wise, is to have lots and lots of observations. Observations on just a few people are just anecdotal evidence. (“I put the crystal on my head and it healed me!”)
The fact that the authors used millions of people to try to find relationships pretty conclusively establishes there is NO relationship between horoscope and…well, anything, really.
Starting without a hypothesis or specific relationship between your variables is just fishing. It’s really just not good science. And these authors have proven that with a very nice piece of science work indeed!
One other news release from AAAS: Run for Science! Since Rebecca urged you to write Oprah, I thought I would urge you to run for office.