lower Compat minimum version (#45)

* lower Compat minimum version

0.17 needed for new-style abstract type

* remove unneeded compat const

would be needed if parametric (in Compat 0.18), but it isn't here

* remove unneeded at-compat in bivariate.jl

ditto, should squash this

Author: tkelman

REQUIRE

@@ -1,5 +1,5 @@
 julia 0.4
-Compat 0.19.0
+Compat 0.17
 StatsBase
 Distributions
 Optim

src/bivariate.jl

@@ -85,7 +85,7 @@ function conv(k::BivariateKDE, dist::Tuple{UnivariateDistribution,UnivariateDist
     BivariateKDE(k.x, k.y, dens)
 end
 
-@compat const BivariateDistribution = Union{MultivariateDistribution,Tuple{UnivariateDistribution,UnivariateDistribution}}
+const BivariateDistribution = Union{MultivariateDistribution,Tuple{UnivariateDistribution,UnivariateDistribution}}
 
 function kde(data::Tuple{RealVector, RealVector}, midpoints::Tuple{Range, Range}, dist::BivariateDistribution)
     k = tabulate(data,midpoints)

src/univariate.jl

@@ -8,7 +8,7 @@ end
 kernel_dist(::Type{Normal},w::Real) = Normal(0.0,w)
 kernel_dist(::Type{Uniform},w::Real) = (s = 1.7320508075688772*w; Uniform(-s,s))
 
-@compat const LocationScale = Union{Laplace,Logistic,SymTriangularDist}
+const LocationScale = Union{Laplace,Logistic,SymTriangularDist}
 kernel_dist{D}(::Type{D},w::Real) = (s = w/std(D(0.0,1.0)); D(0.0,s))