Enforce naming consistencies for Chisq & Cosine

src/univariate/continuous/chisq.jl

@@ -1,59 +1,59 @@
 immutable Chisq <: ContinuousUnivariateDistribution
-    df::Float64
+    ν::Float64
 
-    function Chisq(k::Real)
-        k > zero(k) || error("The degree of freedom k must be positive")
-        @compat new(Float64(k))
+    function Chisq(ν::Real)
+        ν > zero(ν) ||
+            throw(ArgumentError("Chisq: ν must be positive."))
+        @compat new(Float64(ν))
     end
 end
 
 @distr_support Chisq 0.0 Inf
 
 #### Parameters
 
-dof(d::Chisq) = d.df
-params(d::Chisq) = (d.df,)
+dof(d::Chisq) = d.ν
+params(d::Chisq) = (d.ν,)
 
 
 #### Statistics
 
-mean(d::Chisq) = dof(d)
+mean(d::Chisq) = d.ν
 
-var(d::Chisq) = 2.0 * dof(d)
+var(d::Chisq) = 2.0 * d.ν
 
-skewness(d::Chisq) = sqrt(8.0 / dof(d))
+skewness(d::Chisq) = sqrt(8.0 / d.ν)
 
-kurtosis(d::Chisq) = 12.0 / dof(d)
+kurtosis(d::Chisq) = 12.0 / d.ν
 
-mode(d::Chisq) = d.df > 2.0 ? d.df - 2.0 : 0.0
+mode(d::Chisq) = d.ν > 2.0 ? d.ν - 2.0 : 0.0
 
 function median(d::Chisq; approx::Bool=false)
     if approx
-        k = dof(d)
-        return k * (1.0 - 2.0 / (9.0 * k))^3
+        return d.ν * (1.0 - 2.0 / (9.0 * d.ν))^3
     else
         return quantile(d, 0.5)
     end
 end
 
 function entropy(d::Chisq)
-    hk = 0.5 * dof(d)
-    hk + logtwo + lgamma(hk) + (1.0 - hk) * digamma(hk)
+    hν = 0.5 * d.ν
+    hν + logtwo + lgamma(hν) + (1.0 - hν) * digamma(hν)
 end
 
 
 #### Evaluation
 
-@_delegate_statsfuns Chisq chisq df
+@_delegate_statsfuns Chisq chisq ν
 
-mgf(d::Chisq, t::Real) = (1.0 - 2.0 * t)^(-dof(d) / 2.0)
+mgf(d::Chisq, t::Real) = (1.0 - 2.0 * t)^(-d.ν * 0.5)
 
-cf(d::Chisq, t::Real) = (1.0 - 2.0 * im * t)^(-dof(d) / 2.0)
+cf(d::Chisq, t::Real) = (1.0 - 2.0 * im * t)^(-d.ν * 0.5)
 
-gradlogpdf(d::Chisq, x::Float64) =  x >= 0.0 ? (dof(d) / 2.0 - 1) / x - 0.5 : 0.0
+gradlogpdf(d::Chisq, x::Float64) =  x > 0.0 ? (d.ν * 0.5 - 1) / x - 0.5 : 0.0
 
 
 #### Sampling
 
 _chisq_rand(ν::Float64) = StatsFuns.Rmath.chisqrand(ν)
-rand(d::Chisq) = _chisq_rand(d.df)
+rand(d::Chisq) = _chisq_rand(d.ν)

src/univariate/continuous/cosine.jl

@@ -5,26 +5,26 @@
 
 immutable Cosine <: ContinuousUnivariateDistribution
     μ::Float64
-    s::Float64
+    σ::Float64
 
-    function Cosine(μ::Real, s::Real)
-        s > 0.0 || error("s must be positive.")
-        @compat new(Float64(μ), Float64(s))
+    function Cosine(μ::Real, σ::Real)
+        σ > 0.0 || error("Cosine: σ must be positive.")
+        @compat new(Float64(μ), Float64(σ))
     end
 
-    @compat Cosine(μ::Real) = new(Float64(μ), 1.0)
+    Cosine(μ::Real) = @compat new(Float64(μ), 1.0)
     Cosine() = new(0.0, 1.0)
 end
 
-@distr_support Cosine d.μ - d.s d.μ + d.s
+@distr_support Cosine d.μ - d.σ d.μ + d.σ
 
 
 #### Parameters
 
 location(d::Cosine) = d.μ
-scale(d::Cosine) = d.s
+scale(d::Cosine) = d.σ
 
-params(d::Cosine) = (d.μ, d.s)
+params(d::Cosine) = (d.μ, d.σ)
 
 
 #### Statistics
@@ -35,8 +35,7 @@ median(d::Cosine) = d.μ
 
 mode(d::Cosine) = d.μ
 
-const _cosined_varcoef = 0.13069096604865779  # 1 / 3 - 2 / π^2
-var(d::Cosine) = d.s^2 * _cosined_varcoef
+var(d::Cosine) = d.σ^2 * 0.13069096604865779  # 0.130... = 1/3 - 2 / π^2
 
 skewness(d::Cosine) = 0.0
 
@@ -47,9 +46,8 @@ kurtosis(d::Cosine) = -0.59376287559828102362
 
 function pdf(d::Cosine, x::Float64)
     if insupport(d, x)
-        μ, s = params(d)
-        z = (x - μ) / s
-        return (1.0 + cospi(z)) / (2 * s)
+        z = (x - d.μ) / d.σ
+        return (1.0 + cospi(z)) / (2 * d.σ)
     else
         return 0.0
     end
@@ -58,14 +56,12 @@ end
 logpdf(d::Cosine, x::Float64) = insupport(d, x) ? log(pdf(d, x)) : -Inf
 
 function cdf(d::Cosine, x::Float64)
-    μ, s = params(d)
-    z = (x - μ) / s
+    z = (x - d.μ) / d.σ
     0.5 * (1.0 + z + sinpi(z) * invπ)
 end
 
 function ccdf(d::Cosine, x::Float64)
-    μ, s = params(d)
-    nz = (μ - x) / s
+    nz = (d.μ - x) / d.σ
     0.5 * (1.0 + nz + sinpi(nz) * invπ)
 end