Closed KronosTheLate closed 2 years ago
I just check the wiki article: https://en.wikipedia.org/wiki/P-value There, it is stated that
I can see that this definition is in agreement with the result provided in this package, and thereby close this issue.
This is however a bit if a sticky point, as this stackexchange discussion elaborates on.
from
f.jl
:How does this implementation account for the fact that a
FDist
is not symmetic, and that thecdf
up to a point of a given probability density is not the same as thecdf
after the next point of equal probability density?Example: I want to test if two distributions have the same variance. I get very different results from the implemented
pvalue
, and the one I calculated in the way thought in my class:There is a pretty huge difference in some cases, as seen here. Have I misunderstood something, or is the current implementation of
pvalue
forFDist
wrong? If it is wrong, is it different for other non-symmetric distributions?Code without outputs for easier copying: