When computing the correlation between a matrix and a vector, corspearman returns a 10x1 matrix while corkendall returns a 10-element vector. There should be consistent return types between these two functions.
Minimum example:
julia> z = rand(100, 100)
julia> x = rand(100)
julia> corspearman(z, x)
10×1 Matrix{Float64}:
[...]
julia corkendall(z, x)
10-element Vector{Float64}:
[...]
When switching the input order (corkendall(x, z) or corspearman(x, z)), the return type for both is a 1x10 Matrix, so I believe that corkendall(Matrix, Vector) should return a Matrix with 1 column.