NCKA poster Northern Boy posted this. I think he's right. If all attacks had involved Hobies or none had, the trend would be convincing:
Quote:
I'm not so sure about that. Statistics is an exact science, but the design and interpretation of statistical tests is not.
There are a lot of factors to be considered in the latter - the importance of those depends on the strength of the effect - to give a simplistic example - if the 12 attacks had all been on Hobies, then I don't think anyone would be worrying about whether Hobies spend more time on the water or what color they were.
However, as it is, the descriptive statistics available are indeed pretty muddled (I had not seen those before, they are very informative) and thus the small sample size is a problem, as are the other, potentially confounding, factors like the number of Hobies vs paddle etc. Hopefully not a 'problem' that will be fixed anytime soon.
NB
|