Saturday, February 5, 2011

Punxsutawney Phil Stats (Courtesy of My Sister Laurie)

Here's a good Facebook discussion by my sister and her friend on the statistical accuracy of Puxatony Phil:

Sister: Just spent the last few minutes calculating Puxatony Phil's statistical accuracy. Of 114 years of forecasts, he has only been 39% accurate. Unfortunately, this means his shadow is not at chance, but instead is actually statistically *inaccurate* at p < 0.02 two-tailed. Based on his prognosis today, this makes me sad.

Sister's Friend: I'm not sure your statistical approach is valid, Laurie. I don't think long and short winters are equally likely, and Phil seems to have a bias towards seeing his shadow. We need to do a signal detection analysis.

Sister: Here are the actual stats in case anyone would like to run their own analyses: Sees Shadow- Phil was right 37 of 99 times. No Shadow- Phil was right 7 of 15 times

Sister's Friend: So early springs are more likely than long winters (69/114), and Phil tends to see his shadow more often than not (99/114). Since a shadow is supposed to predict a long winter, this accounts for the negative relationship. But that doesn't make Phil a useful anti-guide: d' = -0.4, which is pretty near chance.

Sister: Great work, [Friend]. Also nice to see that early springs are slightly though not statistically more likely than late winters.

For reference: http://www.lifeslittlemysteries.com/punxsutawney-phil-weather-prediction-accuracy-1292/

4 comments:

  1. That's awesome. This conversation reminded me of my graduate school days learning to use statistics and playing with data on a program called SPSS (I have a Masters and BA in applied psychology). I'd love to see that raw data on Puxatony's Phils and mess around with it.

    ReplyDelete
  2. Oh, and the lifeslittlemysteries is an awesome website I haven't seen before! Bookmarked as well.

    ReplyDelete
  3. Thanks for the kind words, Joey. It always makes my day to hear when people enjoy what I write, even if the only people who read it have advanced science degrees :-)

    ReplyDelete
  4. Aw man, I think ANYBODY would benefit from this stuff if they just gave it a chance. I guess we're the few people who didn't get fully intimidated by this logic/math stuff. [There's a Fermi problem in there somewhere ;-)]

    In grade school I had horrible math teachers who made the subject frightening to me and it was only in college that I learned to appreciate this kind of stuff again.

    ReplyDelete