problem with Wilcoxon signed rank test

[tcporco]

HypothesisTests v0.8.0, Julia 1.0.1 MacOS

1. the package is working:

julia> SignedRankTest([0.228,-5.58,-2.04,-1.43,1.80,2.52,0.17,3.07,1.37,0.86,0.15,-3.36,0.14,0.23,-0.89])
Exact Wilcoxon signed rank test
-------------------------------
Population details:
    parameter of interest:   Location parameter (pseudomedian)
    value under h_0:         0
    point estimate:          0.17
    95% confidence interval: (-1.465, 1.015)

Test summary:
    outcome with 95% confidence: fail to reject h_0
    two-sided p-value:           0.8469

Details:
    number of observations:      15
    Wilcoxon rank-sum statistic: 64.0
    rank sums:                   [64.0, 56.0]
    adjustment for ties:         0.0

2. Except when it isn't:

julia> SignedRankTest([-0.228,-5.58,-2.04,-1.43,1.80,2.52,0.17,3.07,1.37,0.86,0.15,-3.36,0.14,0.23,-0.89])

p-values must be in [0.,1.]

Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] Type at /Users/travisporco/.julia/packages/StatsBase/NzjNi/src/statmodels.jl:385 [inlined]
 [3] show(::IOContext{Base.GenericIOBuffer{Array{UInt8,1}}}, ::ExactSignedRankTest{Float64}) at /Users/travisporco/.julia/packages/HypothesisTests/M3Ysg/src/HypothesisTests.jl:122
 [4] show at ./sysimg.jl:194 [inlined]
 [5] limitstringmime(::MIME{Symbol("text/plain")}, ::ExactSignedRankTest{Float64}) at /Users/travisporco/.julia/packages/IJulia/0cLgR/src/inline.jl:36
 [6] display_mimestring(::MIME{Symbol("text/plain")}, ::ExactSignedRankTest{Float64}) at /Users/travisporco/.julia/packages/IJulia/0cLgR/src/execute_request.jl:73
 [7] display_dict(::ExactSignedRankTest{Float64}) at /Users/travisporco/.julia/packages/IJulia/0cLgR/src/execute_request.jl:101
 [8] #invokelatest#1 at ./essentials.jl:697 [inlined]
 [9] invokelatest at ./essentials.jl:696 [inlined]
 [10] execute_request(::ZMQ.Socket, ::IJulia.Msg) at /Users/travisporco/.julia/packages/IJulia/0cLgR/src/execute_request.jl:256
 [11] #invokelatest#1 at ./essentials.jl:697 [inlined]
 [12] invokelatest at ./essentials.jl:696 [inlined]
 [13] eventloop(::ZMQ.Socket) at /Users/travisporco/.julia/packages/IJulia/0cLgR/src/eventloop.jl:8
 [14] (::getfield(IJulia, Symbol("##12#15")))() at ./task.jl:259

3. Correct answer to problem:

R> wilcox.test(c(-0.228,-5.58,-2.04,-1.43,1.80,2.52,0.17,3.07,1.37,0.86,0.15,-3.36,0.14,0.23,-0.89))

	Wilcoxon signed rank test

data:  c(-0.228, -5.58, -2.04, -1.43, 1.8, 2.52, 0.17, 3.07, 1.37, 0.86, 0.15, -3.36, 0.14, 0.23, -0.89)
V = 60, p-value = 1
alternative hypothesis: true location is not equal to 0

[nalimilan]

FWIW, the p-value is really out of the correct range:

julia> pvalue(SignedRankTest([-0.228,-5.58,-2.04,-1.43,1.80,2.52,0.17,3.07,1.37,0.86,0.15,-3.36,0.14,0.23,-0.89]))
1.0220336914062502

[ararslan]

Wow, that is incredibly probable