3 ms·

It’s a good question. Is there a statistical test that can be / has been done with the data? Are the numbers between Pfizer and Moderna statistically different

• contravariant 5y ago
  For things that happen with a small probability over lots of trials you can generally assume they're approximately Poisson distributed. Which on its own isn't too helpful but then you know that the standard deviation is equal to the square root of the average. In this case both values bracket 100 so the square root should be around 10 making this a difference of roughly 5 standard deviations. This isn't a Gaussian distribution so the 5-sigma limit used in physics can't be applied directly, but it's still a pretty big difference, especially for medical data. You could (and in some cases should) do the work to get a proper number for the significance but this back of the napkin estimate suggests it's a statistically meaningful difference. And this being a calculation that can be done at the top of one's head, it is useful when the news reports on a change in some statistic that isn't anywhere close to significant.