How reliable are Phil's predictions?
After a too-long internet search, the best historical weather data I was able to find (I'm happy to do something else if you can find better data) was monthly average temperatures for Portland, OR from 1941 on. Since Phil predicts "early spring" or "six more weeks of winter," I gave February a weight of 4 and March a weight of 2 (number of weeks of those six weeks) and took the average. Not the best method of course, but the best I have available. There were no predictions for 1941 and 1942 and no weather data for 2007 and 2008, so I had 63 data points.
Phil predicts more winter:
Number of times - 53
Mean temperature - 44.38
Variance - 4.94
Phil predicts early spring:
Number of times - 10
Mean temperature - 44.96
Variance - 4.90
Two-sample t-test:
T-statistic: -0.760 with 61 degrees of freedom
One-sided p-value: 0.225
***************************************************
Translation: We say these results are so extreme they would "only" happen one in four times by chance. Which isn't sufficient evidence to claim that there's any relationship between Groundhog's Day predictions and actual outcomes.
Sorry to eat away at your childhood :)
Subscribe to:
Post Comments (Atom)
Posts
I like seeing stupid holidays taken down a peg, especially when they are so clearly meaningless. Thanks, Chris.
ReplyDelete