Fix deprecations (#110)

src/HypothesisTests.jl

@@ -33,7 +33,7 @@ using Rmath: pwilcox, psignrank
 import StatsBase.confint
 
 export testname, pvalue, confint
-@compat abstract type HypothesisTest end
+abstract type HypothesisTest end
 
 check_same_length(x::AbstractVector, y::AbstractVector) = if length(x) != length(y)
     throw(DimensionMismatch("Vectors must be the same length"))
@@ -99,7 +99,7 @@ function check_alpha(alpha::Float64)
 end
 
 # Pretty-print
-function Base.show{T<:HypothesisTest}(io::IO, test::T)
+function Base.show(io::IO, test::T) where T<:HypothesisTest
     println(io, testname(test))
     println(io, repeat("-", length(testname(test))))
 
@@ -136,12 +136,12 @@ function Base.show{T<:HypothesisTest}(io::IO, test::T)
 end
 
 # parameter of interest: name, value under h0, point estimate
-population_param_of_interest{T<:HypothesisTest}(test::T) = ("not implemented yet", NaN, NaN)
+population_param_of_interest(test::T) where {T<:HypothesisTest} = ("not implemented yet", NaN, NaN)
 
 # is the test one- or two-sided
 default_tail(test::HypothesisTest) = :undefined
 
-function show_params{T<:HypothesisTest}(io::IO, test::T, ident="")
+function show_params(io::IO, test::T, ident="") where T<:HypothesisTest
     fieldidx = find(Bool[t<:Number for t in T.types])
     if !isempty(fieldidx)
         lengths = [length(string(T.names[i])) for i in fieldidx]

src/anderson_darling.jl

@@ -2,12 +2,12 @@
 
 export OneSampleADTest, KSampleADTest
 
-@compat abstract type ADTest <: HypothesisTest end
+abstract type ADTest <: HypothesisTest end
 
 ## ONE SAMPLE AD-TEST
 ### http://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm
 
-function adstats{T<:Real}(x::AbstractVector{T}, d::UnivariateDistribution)
+function adstats(x::AbstractVector{T}, d::UnivariateDistribution) where T<:Real
     n = length(x)
     y = sort(x)
     μ = mean(y)
@@ -23,7 +23,7 @@ function adstats{T<:Real}(x::AbstractVector{T}, d::UnivariateDistribution)
     (n, μ, σ, A²)
 end
 
-immutable OneSampleADTest <: ADTest
+struct OneSampleADTest <: ADTest
     n::Int      # number of observations
     μ::Float64  # sample mean
     σ::Float64  # sample std
@@ -39,7 +39,7 @@ is not drawn from `d`.
 
 Implements: [`pvalue`](@ref)
 """
-function OneSampleADTest{T<:Real}(x::AbstractVector{T}, d::UnivariateDistribution)
+function OneSampleADTest(x::AbstractVector{T}, d::UnivariateDistribution) where T<:Real
     OneSampleADTest(adstats(x, d)...)
 end
 
@@ -72,7 +72,7 @@ function pvalue(x::OneSampleADTest)
 end
 
 ## K-SAMPLE ANDERSON DARLING TEST
-immutable KSampleADTest <: ADTest
+struct KSampleADTest <: ADTest
     k::Int        # number of samples
     n::Int        # number of observations
     σ::Float64   # variance A²k
@@ -96,7 +96,7 @@ Implements: [`pvalue`](@ref)
   * F. W. Scholz and M. A. Stephens, K-Sample Anderson-Darling Tests, Journal of the
     American Statistical Association, Vol. 82, No. 399. (Sep., 1987), pp. 918-924.
 """
-function KSampleADTest{T<:Real}(xs::AbstractVector{T}...; modified=true)
+function KSampleADTest(xs::AbstractVector{T}...; modified=true) where T<:Real
     KSampleADTest(a2_ksample(xs, modified)...)
 end
 

src/augmented_dickey_fuller.jl

@@ -24,7 +24,7 @@
 
 export ADFTest
 
-immutable ADFTest <: HypothesisTest
+struct ADFTest <: HypothesisTest
     n               ::Int               # number of observations
     deterministic   ::Symbol            # deterministic terms included in regression
     lag             ::Int               # number of lags in test statistic
@@ -58,7 +58,7 @@ External links
 
   * [Augmented Dickey-Fuller test on Wikipedia](https://en.wikipedia.org/wiki/Augmented_Dickey–Fuller_test)
 """
-function ADFTest{T<:Real}(y::AbstractVector{T}, deterministic::Symbol, lag::Int)
+function ADFTest(y::AbstractVector{T}, deterministic::Symbol, lag::Int) where T<:Real
 
     nobs = length(y)
     Δy = diff(y)

src/binomial.jl

@@ -26,7 +26,7 @@ export BinomialTest, SignTest
 
 ## BINOMIAL TEST
 
-immutable BinomialTest <: HypothesisTest
+struct BinomialTest <: HypothesisTest
     p::Float64
     x::Int
     n::Int
@@ -176,7 +176,7 @@ end
 
 ## SIGN TEST
 
-immutable SignTest <: HypothesisTest
+struct SignTest <: HypothesisTest
     median::Float64
     x::Int
     n::Int
@@ -193,9 +193,9 @@ hypothesis that the median is not equal to `median`.
 
 Implements: [`pvalue`](@ref), [`confint`](@ref)
 """
-SignTest{T<:Real}(x::AbstractVector{T}, median::Real=0) =
+SignTest(x::AbstractVector{T}, median::Real=0) where {T<:Real} =
     SignTest(median, sum(x .> median), sum(x .!= median), sort(x))
-SignTest{T<:Real, S<:Real}(x::AbstractVector{T}, y::AbstractVector{S}) =
+SignTest(x::AbstractVector{T}, y::AbstractVector{S}) where {T<:Real, S<:Real} =
     SignTest(x - y, 0.0)
 
 testname(::SignTest) = "Sign Test"

src/box_test.jl

@@ -26,7 +26,7 @@ export BoxPierceTest, LjungBoxTest
 
 # Box-Pierce test
 
-immutable BoxPierceTest <: HypothesisTest
+struct BoxPierceTest <: HypothesisTest
     y::Vector{Float64}  # time series vector
     n::Int              # number of observations
     lag::Int            # number of lags in test statistic
@@ -49,7 +49,7 @@ External links
 
 * [Box-Pierce test on Wikipedia](https://en.wikipedia.org/wiki/Ljung–Box_test#Box-Pierce_test)
 """
-function BoxPierceTest{T<:Real}(y::AbstractVector{T}, lag::Int, dof::Int=0)
+function BoxPierceTest(y::AbstractVector{T}, lag::Int, dof::Int=0) where T<:Real
     if dof>=lag
         throw(ArgumentError("Number of lags has to be larger than degrees of" *
         " freedom correction"))
@@ -75,7 +75,7 @@ pvalue(x::BoxPierceTest) = pvalue(Chisq(x.lag-x.dof), x.Q; tail=:right)
 
 #Ljung-Box test
 
-immutable LjungBoxTest <: HypothesisTest
+struct LjungBoxTest <: HypothesisTest
     y::Vector{Float64}  # time series vector
     n::Int              # number of observations
     lag::Int            # number of lags in test statistic
@@ -98,7 +98,7 @@ External links
 
 * [Ljung-Box test on Wikipedia](https://en.wikipedia.org/wiki/Ljung–Box_test)
 """
-function LjungBoxTest{T<:Real}(y::AbstractVector{T}, lag::Int, dof::Int=0)
+function LjungBoxTest(y::AbstractVector{T}, lag::Int, dof::Int=0) where T<:Real
     if dof>=lag
         throw(ArgumentError("Number of lags has to be larger than degrees of" *
         " freedom correction"))

src/breusch_godfrey.jl

@@ -24,7 +24,7 @@
 
 export BreuschGodfreyTest
 
-immutable BreuschGodfreyTest <: HypothesisTest
+struct BreuschGodfreyTest <: HypothesisTest
     n::Int              # number of observations
     lag::Int            # number of lags in test statistic
     BG::Float64         # test statistic
@@ -46,8 +46,8 @@ External links
 
 * [Breusch-Godfrey test on Wikipedia](https://en.wikipedia.org/wiki/Breusch–Godfrey_test)
 """
-function BreuschGodfreyTest{T<:Real}(xmat::AbstractArray{T}, e::AbstractVector{T},
-                                     lag::Int, start0::Bool=true)
+function BreuschGodfreyTest(xmat::AbstractArray{T}, e::AbstractVector{T},
+                            lag::Int, start0::Bool=true) where T<:Real
     n = size(e,1)
     elag = zeros(Float64,n,lag)
     for ii = 1:lag  # construct lagged residuals

src/circular.jl

@@ -34,17 +34,17 @@ export RayleighTest, FisherTLinearAssociation, JammalamadakaCircularCorrelation
 
 ## RAYLEIGH TEST OF UNIFORMITY AGAINST AN UNSPECIFIED UNIMODAL ALTERNATIVE
 
-immutable RayleighTest <: HypothesisTest
+struct RayleighTest <: HypothesisTest
     Rbar::Float64 # mean resultant length
     n::Int        # number of observations
 end
-function RayleighTest{S <: Complex}(samples::Vector{S})
+function RayleighTest(samples::Vector{S}) where S <: Complex
     s = Float64(abs(sum(samples./abs(samples))))
     n = length(samples)
     Rbar = s/n
     RayleighTest(Rbar, n)
 end
-function RayleighTest{S <: Real}(samples::Vector{S})
+function RayleighTest(samples::Vector{S}) where S <: Real
     s = Float64(abs(sum(exp, im * samples)))
     n = length(samples)
     Rbar = s/n
@@ -68,14 +68,14 @@ end
 
 ## N.I. FISHER'S TEST OF T-LINEAR CIRCULAR-CIRCULAR ASSOCIATION
 
-immutable FisherTLinearAssociation{S <: Real, T <: Real} <: HypothesisTest
+struct 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::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::Union{Bool,Void})
+function FisherTLinearAssociation(theta::Vector{Stheta},
+        phi::Vector{Sphi}, uniformly_distributed::Union{Bool,Void}) where {Stheta <: Real, Sphi <: Real}
     check_same_length(theta, phi)
 
     A = sum(cos.(theta).*cos.(phi))
@@ -93,8 +93,8 @@ function FisherTLinearAssociation{Stheta <: Real, Sphi <: Real}(theta::Vector{St
     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
-FisherTLinearAssociation{S <: Real, T <: Real}(theta::Vector{S},
-        phi::Vector{T}) = FisherTLinearAssociation(theta, phi, nothing)
+FisherTLinearAssociation(theta::Vector{S},
+phi::Vector{T}) where {S <: Real, T <: Real} = FisherTLinearAssociation(theta, phi, nothing)
 
 testname(::FisherTLinearAssociation) =
     "T-linear test of circular-circular association"
@@ -188,11 +188,11 @@ end
 
 ## JAMMALAMADAKA'S CIRCULAR CORRELATION
 
-immutable JammalamadakaCircularCorrelation <: HypothesisTest
+struct JammalamadakaCircularCorrelation <: HypothesisTest
     r::Float64  # circular-circular correlation coefficient
     Z::Float64  # test statistic
 end
-function JammalamadakaCircularCorrelation{S <: Real, T <: Real}(alpha::Vector{S}, beta::Vector{T})
+function JammalamadakaCircularCorrelation(alpha::Vector{S}, beta::Vector{T}) where {S <: Real, T <: Real}
     check_same_length(alpha, beta)
     # calculate sample mean directions
     alpha_bar = angle(sum(exp, im * alpha))
@@ -225,7 +225,7 @@ pvalue(x::JammalamadakaCircularCorrelation; tail=:both) = pvalue(Normal(), x.Z;
 # Complex numbers
 for fn in (:JammalamadakaCircularCorrelation, :FisherTLinearAssociation)
     @eval begin
-        $(fn){S <: Complex, T <: Complex}(x::Vector{S}, y::Vector{T}) =
+        $(fn)(x::Vector{S}, y::Vector{T}) where {S <: Complex, T <: Complex} =
             $(fn)(angle(x), angle(y))
     end
 end

src/durbin_watson.jl

@@ -24,7 +24,7 @@
 
 export DurbinWatsonTest
 
-immutable DurbinWatsonTest <: HypothesisTest
+struct DurbinWatsonTest <: HypothesisTest
     xmat::Array{Float64}  # regressor matrix
     n::Int                # number of observations
     DW::Float64           # test statistic
@@ -73,8 +73,8 @@ and `:right` (positive serial correlation).
   https://en.wikipedia.org/wiki/Durbin–Watson_statistic
   ](https://en.wikipedia.org/wiki/Durbin–Watson_statistic)
 """
-function DurbinWatsonTest{T<:Real}(xmat::AbstractArray{T}, e::AbstractArray{T};
-    p_compute::Symbol = :ndep)
+function DurbinWatsonTest(xmat::AbstractArray{T}, e::AbstractArray{T};
+    p_compute::Symbol = :ndep) where T<:Real
 
     n = length(e)
     DW = sum(diff(e) .^2) / sum(e .^2)

src/fisher.jl

@@ -51,7 +51,7 @@ Implements: [`pvalue`](@ref), [`confint`](@ref)
     Blaker’s exact tests". Biostatistics, Volume 11, Issue 2, 1 April 2010, Pages 373–374,
     [link](https://doi.org/10.1093/biostatistics/kxp050)
 """
-immutable FisherExactTest <: HypothesisTest
+struct FisherExactTest <: HypothesisTest
     # Format:
     # X1  X2
     # Y1  a  b

src/jarque_bera.jl

@@ -24,7 +24,7 @@
 
 export JarqueBeraTest
 
-immutable JarqueBeraTest <: HypothesisTest
+struct JarqueBeraTest <: HypothesisTest
     n::Int         # number of observations
     JB::Float64    # test statistic
     skew::Float64  # skewness
@@ -52,7 +52,7 @@ nominal levels up to about 3% and under-sized for larger nominal levels (Mantalo
 
   * [Jarque-Bera test on Wikipedia](https://en.wikipedia.org/wiki/Jarque–Bera_test)
 """
-function JarqueBeraTest{T<:Real}(y::AbstractVector{T})
+function JarqueBeraTest(y::AbstractVector{T}) where T<:Real
     n = length(y)
     M = Base.promote_op(/, T, typeof(n))
     m1r = m2r = m3r = m4r = zero(M)

src/kolmogorov_smirnov.jl

@@ -26,17 +26,17 @@ export
     ExactOneSampleKSTest,
     ApproximateOneSampleKSTest, ApproximateTwoSampleKSTest
 
-@compat abstract type KSTest <: HypothesisTest end
-@compat abstract type ApproximateKSTest <: KSTest end
-@compat abstract type ExactKSTest <: KSTest end
+abstract type KSTest <: HypothesisTest end
+abstract type ApproximateKSTest <: KSTest end
+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
 default_tail(test::KSTest) = :both
 
 ## ONE SAMPLE KS-TEST
 
 # compute supremum of differences between target and empirical cdf before and after the jump of the empirical cdf.
-function ksstats{T<:Real}(x::AbstractVector{T}, d::UnivariateDistribution)
+function ksstats(x::AbstractVector{T}, d::UnivariateDistribution) where T<:Real
     n = length(x)
     cdfs = cdf(d, sort(x))
     δp = maximum((1:n) / n - cdfs)
@@ -47,7 +47,7 @@ end
 
 ### EXACT KOLMOGOROV SMIRNOV TEST
 
-immutable ExactOneSampleKSTest <: ExactKSTest
+struct ExactOneSampleKSTest <: ExactKSTest
     n::Int      # number of observations
     δ::Float64  # supremum of CDF differences
     δp::Float64 # supremum of the positive CDF differences
@@ -63,7 +63,7 @@ sample is not drawn from `d`.
 
 Implements: [`pvalue`](@ref)
 """
-function ExactOneSampleKSTest{T<:Real}(x::AbstractVector{T}, d::UnivariateDistribution)
+function ExactOneSampleKSTest(x::AbstractVector{T}, d::UnivariateDistribution) where T<:Real
     if length(x) > length(unique(x))
         warn("This test is inaccurate with ties")
     end
@@ -91,7 +91,7 @@ end
 
 ### APPROXIMATE KOLMOGOROV SMIRNOV TEST
 
-immutable ApproximateOneSampleKSTest <: ApproximateKSTest
+struct ApproximateOneSampleKSTest <: ApproximateKSTest
     n::Int      # number of observations
     δ::Float64  # supremum of CDF differences
     δp::Float64 # supremum of the positive CDF differences
@@ -107,7 +107,7 @@ that the sample is not drawn from `d`.
 
 Implements: [`pvalue`](@ref)
 """
-function ApproximateOneSampleKSTest{T<:Real}(x::AbstractVector{T}, d::UnivariateDistribution)
+function ApproximateOneSampleKSTest(x::AbstractVector{T}, d::UnivariateDistribution) where T<:Real
     if length(x) > length(unique(x))
         warn("This test is inaccurate with ties")
     end
@@ -139,7 +139,7 @@ end
 
 ### APPROXIMATE KOLMOGOROV SMIRNOV TEST
 
-immutable ApproximateTwoSampleKSTest <: ApproximateKSTest
+struct ApproximateTwoSampleKSTest <: ApproximateKSTest
     n_x::Int    # number of observations
     n_y::Int    # number of observations
     δ::Float64  # supremum of CDF differences
@@ -161,7 +161,7 @@ Implements: [`pvalue`](@ref)
   * [Approximation of one-sided test (Encyclopedia of Mathematics)
     ](https://www.encyclopediaofmath.org/index.php/Kolmogorov-Smirnov_test)
 """
-function ApproximateTwoSampleKSTest{T<:Real, S<:Real}(x::AbstractVector{T}, y::AbstractVector{S})
+function ApproximateTwoSampleKSTest(x::AbstractVector{T}, y::AbstractVector{S}) where {T<:Real, S<:Real}
     n_x, n_y = length(x), length(y)
     if n_x+n_y > length(unique([x; y]))
         warn("This test is inaccurate with ties")
@@ -192,7 +192,7 @@ function pvalue(x::ApproximateTwoSampleKSTest; tail=:both)
 end
 
 # compute supremum of differences between empirical cdfs.
-function ksstats{T<:Real, S<:Real}(x::AbstractVector{T}, y::AbstractVector{S})
+function ksstats(x::AbstractVector{T}, y::AbstractVector{S}) where {T<:Real, S<:Real}
     n_x, n_y = length(x), length(y)
     sort_idx = sortperm([x; y])
     pdf_diffs = [ones(n_x)/n_x; -ones(n_y)/n_y][sort_idx]