Saturday, January 29, 2011

Not even wrong.

No, I'm not talking about Peter Woit's website or Wolfgang Pauli.  Instead, I mean this article, which shows that Allstate Insurance apparently thinks that it's meaningful to look at car accident risk as a function of the astrological sign of the driver.  Astrology?  A major company using astrology?  We're supposed to believe that there is a statistically meaningful correlation between the time of the year you're born and your driving ability?  This is why there is a crying need for math and science literacy.

12 comments:

Schlupp said...

The sad thing is that one of the twelve sign *has* to come out as the one with the most accidents, even though it's just by chance.

CarlBrannen said...

They're claiming the data gives 700% more accidents for one of the signs, this over a few hundred thousand data points. That's more than the usual 5 sigma needed to show an effect.

Douglas Natelson said...

Carl, I just don't believe it. There is no rational reason why people born in a particular month should be 7 times more likely to have accidents. If that's real, and that's a huge "if", I guarantee that it has nothing to do with the positions of the stars in the sky.

Anonymous said...

Maybe there's a variation between the types of parents that plan to have their children during certain times of the year. And perhaps, for whatever reason, parents who plan a certain month of birth for their kids may have certain characteristics that cause them to parent their children to be more accident-prone. I know it sounds far-fetched, but possibly it is a bit less far-fetched than just the positions in the sky.

Anonymous said...

There in fact is evidence that there are correlations involved with the types of mothers and the months their children are born in.

http://www.stat.columbia.edu/~cook/movabletype/archives/2009/09/who_has_babies.html

But frankly insurance companies are practical businesses; they don't need to explain why, they just need to go with what works.

Ψ*Ψ said...

Very, very glad I'm not buying car insurance from them!

Schlupp said...

Also Carl, With enough hypotheses, you are bound to find an "effect" every now and then, regardless of how rigorously you define your bounds.* With 12 zodiac sign that are supposed to influence every aspect of our lives, you can make up a huge lot of hypotheses.

*) With the usual p-value of 0.05, you are practically guaranteed to find *some* sign to show something "significant" as soon as you test only for two things. With more rigorous bounds, the computer has to mine data a bit longer, but it *will* find something sooner or later.

DanM said...

I notice there is no comment on repeatibility. e.g., does the same trend show up in the 2009 data as in the 2010 data? Or is the winner different each year? C'mon, you'd think an insurance company would employ at least *one* statistician.

CarlBrannen said...

Doug, I also don't believe the data. I think someone fed someone a line of complete bull.

Vincent said...

Assuming that the majority of drivers take lessons and pass the exams close to when they are of legal age, the time of year you take lessons and pass the driving test will be correlated with your sign.

I don't find it completely outrages that say for example the probability that a bad driver will pass the test is higher in summer then in winter. Of cause this doesn't match up with their results, but still I think it's to strong to claim that there is no rational reason why your birth time could correlate with accident rates.

That being said, I would like to see their data, I suspect they most likely are wrong in their conclusions, but I don't think a correlation is impossible though I would suspect such a correlation to come from the seasonal differences in our climate and culture, and have nothing to due with the position of stars (which are however a great way to keep track of seasons if you ever find yourself stranded a couple of thousand of years ago!). .

Anonymous said...

I don't see why this is so far-fetched... if many people in this country read and believe their horoscopes (is that true?), and thus are consistently receiving one of two 'types' of advice- advice to be agressive or advice to be patient, it seems reasonable that by following said advice they would be at different risk to crashing their car

sasa said...

Allstate Retracts Failed Zodiac Joke
http://www.huffingtonpost.com/2011/02/03/allstate-zodiac-joke_n_818415.html