Closed bkamins closed 3 years ago
So a tension in this PR that we have:
julia> using Statistics
julia> z = [1.0, im]
2-element Vector{ComplexF64}:
1.0 + 0.0im
0.0 + 1.0im
julia> var(z)
1.0
julia> cov(z, z)
1.0 + 0.0im
and we have a tension that the identity var(z) === cov(z, z)
does not hold. But probably we should accept it.
Yeah, but AFAICT that's an expected result since the variance is guaranteed to be a real number, contrary to the covariance: https://en.wikipedia.org/wiki/Complex_random_variable#Variance_and_pseudo-variance https://selipot.github.io/talks/lecture2.pdf
Anyway var(z) == cov(z, z)
holds so that's already quite consistent.
an expected result since the variance is guaranteed to be a real number
Yes - that is why I said we should accept it.
Was this actually needed for CI to pass?
Ah - I added it to make the tests pass on 1.0. I will remove it
Codecov Report
97.67% <100.00%> (+97.67%)
Continue to review full report at Codecov.