Merge pull request #88 from JuliaStats/aa/0.6

Updates for Julia 0.6

.travis.yml

@@ -4,7 +4,6 @@ os:
   - linux
   - osx
 julia:
-  - 0.4
   - 0.5
   - nightly
 notifications:

README.md

@@ -3,8 +3,8 @@ HypothesisTests.jl
 
 [![Build Status](https://travis-ci.org/JuliaStats/HypothesisTests.jl.svg?branch=master)](https://travis-ci.org/JuliaStats/HypothesisTests.jl)
 [![Coverage Status](https://coveralls.io/repos/JuliaStats/HypothesisTests.jl/badge.svg?branch=master)](https://coveralls.io/r/JuliaStats/HypothesisTests.jl?branch=master)
-[![HypothesisTests](http://pkg.julialang.org/badges/HypothesisTests_0.4.svg)](http://pkg.julialang.org/?pkg=HypothesisTests)
 [![HypothesisTests](http://pkg.julialang.org/badges/HypothesisTests_0.5.svg)](http://pkg.julialang.org/?pkg=HypothesisTests)
+[![HypothesisTests](http://pkg.julialang.org/badges/HypothesisTests_0.6.svg)](http://pkg.julialang.org/?pkg=HypothesisTests)
 
 This package implements several hypothesis tests in Julia.
 

REQUIRE

@@ -1,7 +1,7 @@
-julia 0.4
+julia 0.5
 Distributions 0.10.0
 Roots
 StatsBase 0.9.0
-Compat 0.8.4
+Compat 0.18.0
 Rmath 0.1.2
 Combinatorics

src/HypothesisTests.jl

@@ -22,7 +22,7 @@
 # OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
 # WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
-VERSION >= v"0.4.0-dev+6521" && __precompile__()
+__precompile__()
 
 module HypothesisTests
 
@@ -33,7 +33,7 @@ using Rmath: pwilcox, psignrank
 import StatsBase.confint
 
 export testname, pvalue, confint
-abstract HypothesisTest
+@compat abstract type HypothesisTest end
 
 check_same_length(x::AbstractVector, y::AbstractVector) = if length(x) != length(y)
     throw(DimensionMismatch("Vectors must be the same length"))

src/anderson_darling.jl

@@ -2,7 +2,7 @@
 
 export OneSampleADTest, KSampleADTest
 
-abstract ADTest <: HypothesisTest
+@compat abstract type ADTest <: HypothesisTest end
 
 ## ONE SAMPLE AD-TEST
 ### http://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm
@@ -101,7 +101,7 @@ function pvalue(x::KSampleADTest)
     @inbounds for i = 1:2, k = 1:5
         A[k,i+1] = A[k,i] * tm[k]
     end
-    f = A \ log(sig)
+    f = A \ log.(sig)
 
     exp(f[1] + f[2]*Tk + f[3]*Tk^2)
 end

src/circular.jl

@@ -39,13 +39,13 @@ immutable RayleighTest <: HypothesisTest
     n::Int        # number of observations
 end
 function RayleighTest{S <: Complex}(samples::Vector{S})
-    s = @compat Float64(abs(sum(samples./abs(samples))))
+    s = Float64(abs(sum(samples./abs(samples))))
     n = length(samples)
     Rbar = s/n
     RayleighTest(Rbar, n)
 end
 function RayleighTest{S <: Real}(samples::Vector{S})
-    s = @compat Float64(abs(sum(exp(im*samples))))
+    s = Float64(abs(sum(exp, im * samples)))
     n = length(samples)
     Rbar = s/n
     RayleighTest(Rbar, n)
@@ -71,24 +71,24 @@ immutable FisherTLinearAssociation{S <: Real, T <: Real} <: HypothesisTest
     rho_t::Float64                              # circular correlation coefficient
     theta::Vector{S}                            # radians of group 1
     phi::Vector{T}                              # radians of group 2
-    uniformly_distributed::@compat(Union{Bool,Void}) # is distribution of theta and phi uniform?
+    uniformly_distributed::Union{Bool,Void}     # is distribution of theta and phi uniform?
 end
 function FisherTLinearAssociation{Stheta <: Real, Sphi <: Real}(theta::Vector{Stheta},
-        phi::Vector{Sphi}, uniformly_distributed::@compat(Union{Bool,Void}))
+        phi::Vector{Sphi}, uniformly_distributed::Union{Bool,Void})
     check_same_length(theta, phi)
 
-    A = sum(cos(theta).*cos(phi))
-    B = sum(sin(theta).*sin(phi))
-    C = sum(cos(theta).*sin(phi))
-    D = sum(sin(theta).*cos(phi))
+    A = sum(cos.(theta).*cos.(phi))
+    B = sum(sin.(theta).*sin.(phi))
+    C = sum(cos.(theta).*sin.(phi))
+    D = sum(sin.(theta).*cos.(phi))
     T = A*B-C*D
 
     # Notation drawn from Fisher, 1993
     n = length(theta)
-    E = sum(cos(2*theta))
-    F = sum(sin(2*theta))
-    G = sum(cos(2*phi))
-    H = sum(sin(2*phi))
+    E = sum(cos, 2*theta)
+    F = sum(sin, 2*theta)
+    G = sum(cos, 2*phi)
+    H = sum(sin, 2*phi)
     rho_t = 4*T/sqrt((n^2 - E^2 - F^2)*(n^2-G^2-H^2))
     FisherTLinearAssociation(rho_t, theta, phi, uniformly_distributed)
 end
@@ -105,19 +105,19 @@ end
 # For large samples, compute the distribution and statistic of T
 function tlinear_Z(x::FisherTLinearAssociation)
     n = length(x.theta)
-    theta_resultant = sum(exp(im*x.theta))
-    phi_resultant = sum(exp(im*x.phi))
+    theta_resultant       = sum(exp, im * x.theta)
+    phi_resultant         = sum(exp, im * x.phi)
     theta_resultant_angle = angle(theta_resultant)
-    phi_resultant_angle = angle(phi_resultant)
-    alpha_2_theta = mean(cos(2*(x.theta-theta_resultant_angle)))
-    beta_2_theta = mean(sin(2*(x.theta-theta_resultant_angle)))
-    alpha_2_phi = mean(cos(2*(x.phi-phi_resultant_angle)))
-    beta_2_phi = mean(sin(2*(x.phi-phi_resultant_angle)))
-    U_theta = (1-alpha_2_theta^2-beta_2_theta^2)/2
-    U_phi = (1-alpha_2_phi^2-beta_2_phi^2)/2
-    V_theta = (abs2(theta_resultant)/n^2)*(1-alpha_2_theta)
-    V_phi = (abs2(phi_resultant)/n^2)*(1-alpha_2_phi)
-    sqrt(n)*U_theta*U_phi*rho_t/sqrt(V_theta*V_phi)
+    phi_resultant_angle   = angle(phi_resultant)
+    alpha_2_theta         = mean(cos, 2 * (x.theta - theta_resultant_angle))
+    beta_2_theta          = mean(sin, 2 * (x.theta - theta_resultant_angle))
+    alpha_2_phi           = mean(cos, 2 * (x.phi - phi_resultant_angle))
+    beta_2_phi            = mean(sin, 2 * (x.phi - phi_resultant_angle))
+    U_theta               = (1 - alpha_2_theta^2 - beta_2_theta^2) / 2
+    U_phi                 = (1-alpha_2_phi^2-beta_2_phi^2)/2
+    V_theta               = (abs2(theta_resultant) / n^2) * (1 - alpha_2_theta)
+    V_phi                 = (abs2(phi_resultant) / n^2) * (1 - alpha_2_phi)
+    return sqrt(n) * U_theta * U_phi * rho_t / sqrt(V_theta * V_phi)
 
     # Alternative computational strategy from Fisher and Lee (1983)
     # a1 = [mean(cos(theta)), mean(cos(phi))]
@@ -141,10 +141,10 @@ function pvalue(x::FisherTLinearAssociation; tail=:both)
         # distribution is uniform, use a permutation test.
 
         # "For n < 25, use a randomisation test based on the quantity T = AB - CD"
-        ct = cos(x.theta)
-        st = sin(x.theta)
-        cp = cos(x.phi)
-        sp = sin(x.phi)
+        ct = cos.(x.theta)
+        st = sin.(x.theta)
+        cp = cos.(x.phi)
+        sp = sin.(x.phi)
         T = sum(ct.*cp)*sum(st.*sp)-sum(ct.*sp)*sum(st.*cp)
         greater = 0
 
@@ -192,16 +192,17 @@ end
 function JammalamadakaCircularCorrelation{S <: Real, T <: Real}(alpha::Vector{S}, beta::Vector{T})
     check_same_length(alpha, beta)
     # calculate sample mean directions
-    alpha_bar = angle(sum(exp(im*alpha)))
-    beta_bar = angle(sum(exp(im*beta)))
-    r = sum(sin(alpha .- alpha_bar).*sin(beta .- beta_bar))/sqrt(sum(sin(alpha .- alpha_bar).^2)*sum(sin(beta .- beta_bar).^2))
+    alpha_bar = angle(sum(exp, im * alpha))
+    beta_bar = angle(sum(exp, im * beta))
+    r = sum(sin.(alpha .- alpha_bar) .* sin.(beta .- beta_bar)) /
+        sqrt(sum(sin.(alpha .- alpha_bar).^2) * sum(sin.(beta .- beta_bar).^2))
 
-    sin2_alpha = sin(alpha .- alpha_bar).^2
-    sin2_beta = sin(beta .- beta_bar).^2
+    sin2_alpha = sin.(alpha .- alpha_bar).^2
+    sin2_beta = sin.(beta .- beta_bar).^2
     lambda_20 = mean(sin2_alpha)
     lambda_02 = mean(sin2_beta)
-    lambda_22 = mean(sin2_alpha.*sin2_beta)
-    Z = sqrt(length(alpha)*lambda_20*lambda_02/lambda_22)*r
+    lambda_22 = mean(sin2_alpha .* sin2_beta)
+    Z = sqrt(length(alpha) * lambda_20 * lambda_02 / lambda_22) * r
 
     JammalamadakaCircularCorrelation(r, Z)
 end

src/common.jl

@@ -56,4 +56,4 @@ function tiedrank_adj!(ord::AbstractVector, v::AbstractArray)
     (ord, tieadj)
 end
 
-tiedrank_adj(v::AbstractArray) = tiedrank_adj!(Array(Float64, length(v)), v)
+tiedrank_adj(v::AbstractArray) = tiedrank_adj!(Vector{Float64}(length(v)), v)

src/kolmogorov_smirnov.jl

@@ -26,9 +26,9 @@ export
     ExactOneSampleKSTest,
     ApproximateOneSampleKSTest, ApproximateTwoSampleKSTest
 
-abstract KSTest <: HypothesisTest
-abstract ApproximateKSTest <: KSTest
-abstract ExactKSTest <: KSTest
+@compat abstract type KSTest <: HypothesisTest end
+@compat abstract type ApproximateKSTest <: KSTest end
+@compat abstract type ExactKSTest <: KSTest end
 
 population_param_of_interest(x::KSTest) = ("Supremum of CDF differences", 0.0, x.δ) # parameter of interest: name, value under h0, point estimate
 

src/kruskal_wallis.jl

@@ -67,7 +67,7 @@ function kwstats{T<:Real}(groups::AbstractVector{T}...)
     C = 1-tieadj/(n^3 - n)
 
     # compute rank sums
-    R_i = Array(Float64, length(groups))
+    R_i = Vector{Float64}(length(groups))
     n_end = 0
     for i=1:length(groups)
         R_i[i] = sum(ranks[n_end+1:n_end+n_i[i]])
@@ -79,4 +79,4 @@ function kwstats{T<:Real}(groups::AbstractVector{T}...)
     H /= C
 
     (H, R_i, C, n_i)
-end
+end