We often heard people telling that there if more births during the full moon because they know somebody who know somebody who know somebody working in a maternity who say that's true.
OK, my precedent sentence is a little bit inflated, but every time I hear this I have serious doubts.
So, is that true or not ?
INSEE made public statistics about birth, so it is possible to try to verify this point. But first, I would say I'm not a statistitian at all, so if you see some mistake there, fell free to report.
I took a four year sample and get this :
|Year||2009||2010||2011||2012||four year average|
|Full moon average||2185||2130||2219||2208||2185.5|
|Deviation from the annual average||0.55%||-3.09%||2.16%||2.27%||0.46%|
Let's forget a few time that a 4 year is not a so representative sample.
There is less than 1% difference between full moon days and the others. If there is people working in maternity who really think there is more births during the full moon it is probably because they only remember the full moon days with more birth than usual because it is more consistent with their beliefs (Or, by chance, this could locally be true).
Go back to the representativeness of our sample. We have to determine margin of error which we are sure it will encompass the result if we made the statement in an infinite number of birth (the infinite number already being an aproximation). The problem is that we have to choose the reliability of this margin of error.
There I choose a 99% confidence interval.
The confidence interval for the average number of birth is 3.37% and for the number of birth during the full moon is 18.24% !
Interesting, isn't it? ;-)
The more interesting point is that a 3 178 930 birth sample is not enough to have a certitude (however we have a good presumption).
The second point is that we often heard about statistics with a margin of error. Which is good but not enough, we must have also the confidence interval !
So, the next time you will hear that 83% of french people +/-3% eat sauerkraut for breakfast, ask if it is a 99%, 98%, 95% or 5 % confidence interval!
No comment yet