How do I extract more than a pvalue HypothesisTests.jl?

Rosejoycrocker · September 30, 2024, 12:12am

Hi,
I’m using HypothesisTests.jl to do a series of 2 Sample Kolmogorov Smirnov tests. This means I’m looping through a large series of data sets and applying the test KS_object = ApproximateTwoSampleKSTest(data_set1, data_set2) . I need to be able to store the KS statistic as I loop through the datasets but I can’t see how to do this. If I look at fieldnames(typeof(KS_object)) , the KS stat is not stored in the fieldnames. I also can’t use the pvalue function as this is not what I want to extract. Any advice appreciated, the docs don’t have any examples for this test.

nilshg · September 30, 2024, 7:50am

You can see what the test is showing you here:

github.com

JuliaStats/HypothesisTests.jl/blob/master/src/kolmogorov_smirnov.jl#L172


      
              ](https://www.encyclopediaofmath.org/index.php/Kolmogorov-Smirnov_test)
          """
          function ApproximateTwoSampleKSTest(x::AbstractVector{T}, y::AbstractVector{S}) where {T<:Real, S<:Real}
              n_x, n_y = length(x), length(y)
          
              ApproximateTwoSampleKSTest(ksstats(x, y)...)
          end
          
          testname(::ApproximateTwoSampleKSTest) = "Approximate two sample Kolmogorov-Smirnov test"
          
          function show_params(io::IO, x::ApproximateTwoSampleKSTest, ident="")
              n = x.n_x*x.n_y/(x.n_x+x.n_y)
              println(io, ident, "number of observations:   [$(x.n_x),$(x.n_y)]")
              println(io, ident, "KS-statistic:              $(sqrt(n)*x.δ)")
          end
          
          function StatsAPI.pvalue(x::ApproximateTwoSampleKSTest; tail=:both)
              n = x.n_x*x.n_y/(x.n_x+x.n_y)
              if tail == :left
                  exp(-2*n*x.δn^2)
              elseif tail == :right

To see:

julia> x = ApproximateTwoSampleKSTest(rand(10), rand(10))
Approximate two sample Kolmogorov-Smirnov test
----------------------------------------------
Population details:
    parameter of interest:   Supremum of CDF differences
    value under h_0:         0.0
    point estimate:          0.5

Test summary:
    outcome with 95% confidence: fail to reject h_0
    two-sided p-value:           0.1641

Details:
    number of observations:   [10,10]
    KS-statistic:              1.118033988749895


julia> n = x.n_x*x.n_y/(x.n_x+x.n_y)
5.0

julia> √(n)*x.δ
1.118033988749895