fisky wrote:
These questions are covid-related and we have some statisticians/medical professionals on this thread so I'm going to ask them.
Why is P ≤ 0.05 statistically significant and P ≥ 0.05 statistically insignificant? I get that there needs to be a cutoff, but why is it a yes/no cutoff rather than a scaled level of significance?
As a layperson, it seems very arbitrary to me that 0.05 is significant and 0.051 is not significant.
It is pretty arbitrary but it's also a widely used convention so a useful shorthand. There are other conventions but the 95% confidence level for "statistically significant" is commonly accepted. You'll see higher or lower confidence levels depending on the consequences of being "statistically wrong". No one, for instance, would accept a 95% confidence level that their car brakes would work.
It's better practice, and this is common in the literature, to explicitly state the confidence levels calculated.