The idea here is that it’s not about the number of heads, it’s that we’ve found even one. A statistical test is a bit like flipping a coin; if it’s heads, it’s significant, if it’s tails, it isn’t. If all we care about is getting a ‘significant’ result, we’ll just keep on flipping until we get a heads.

What the Bonferroni test does is essentially what you are saying; it makes the level required for significance higher proportional to the number of tests we do so that we can’t just keep flipping coins to get the result that we want.

Epidemiologist. Writer. Podcaster. Twitter FB Email