Distributions.jl
Distributions.jl copied to clipboard
`0 <= cdf(truncated(Normal(2.5, 0.2), lower=0.0),x)` not always true.
Followup to #1853
Consider the following:
using Distributions
D = truncated(Normal(2.5, 0.2), lower=0.0)
cdf(D, 3.741058503233821e-17) > 0 # false
cdf(D, 1.4354474178676617e-18) > 0 # false
cdf(D, 8.834854780587132e-18) > 0 # false
I think I am simply hitting catastrophic cancellation in this line.
I understand that such numerical incauracies are always possible at borders, but would that be considered a bug ?