The Guardian’s analysis of the forecasting skill of groundhogs like Punxsutawney Phil (beanie baby version in Photo) and Staten Island Chuck:
“We have taken the behaviour of groundhogs for years going back to 1999. We have used this to calculate the modal behaviour of the groundhogs, that is effectively the 'groundhog consensus' on the matter of winter each year.
Now for the tricky part: how do we measure if the winter ended or continued for six weeks? We have taken snow cover in North America for February of every year to 1999, but this alone does not tell us if the groundhog was right about winter ending. To work this out we need to know how snowy a given February was relative to an average February. We calculated the mean average snowfall for a February in North America using data from the last 10 years, then we subtracted this average from the snow cover for the particular year. This gives us an indication of the severity of the winter for that year relative to the other years.
The conclusion of our little study is that groundhogs have only predicted the length of winter correctly four times in the last 13 years. “
Yes, but only a 13 year sample? Clearly we need a Bigger Dataset:)
Comments