wand bins patch: fix #479

Project.toml

@@ -3,6 +3,7 @@ uuid = "f3b207a7-027a-5e70-b257-86293d7955fd"
 version = "0.14.30"
 
 [deps]
+AbstractFFTs = "621f4979-c628-5d54-868e-fcf4e3e8185c"
 Clustering = "aaaa29a8-35af-508c-8bc3-b662a17a0fe5"
 DataStructures = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
 DataValues = "e7dc6d0d-1eca-5fa6-8ad6-5aecde8b7ea5"

src/StatsPlots.jl

@@ -20,7 +20,7 @@ import Clustering: Hclust, nnodes
 using Interpolations
 import MultivariateStats: MDS, eigvals, projection, principalvars,
                             principalratio, transform
-
+import AbstractFFTs: fft, ifft
 import KernelDensity
 @recipe f(k::KernelDensity.UnivariateKDE) = k.x, k.density
 @recipe f(k::KernelDensity.BivariateKDE) = k.x, k.y, permutedims(k.density)

src/hist.jl

@@ -207,12 +207,12 @@ function bkfe(gcounts, drv, bandwidth, range_x)
     ## Obtain kernel weights
 
     tau = 4 + drv
-    L = min(floor(tau*h/delta), M)
+    L = min(floor(Int, tau*h/delta), M)
 
     lvec = 0:L
     arg = lvec .* delta/h
 
-    kappam = pdf(Normal(),arg)/(h^(drv+1))
+    kappam = pdf.(Normal(),arg)/(h^(drv+1))
     hmold0, hmnew = ones(length(arg)), ones(length(arg))
     hmold1 = arg
 
@@ -227,7 +227,7 @@ function bkfe(gcounts, drv, bandwidth, range_x)
 
     ## Now combine weights and counts to obtain estimate
     ## we need P >= 2L+1L, M: L <= M.
-    P = 2^(ceil(log(M+L+1)/log(2)))
+    P = 2^(ceil(Int, log(M+L+1)/log(2)))
     kappam = [kappam; zeros(P-2*L-1); reverse(kappam[2:end])]
     Gcounts = [gcounts; zeros(P-M)]
     kappam = fft(kappam)

test/runtests.jl

@@ -59,6 +59,11 @@ end # testset
     ]
 end
 
+@testset "Histogram" begin
+    data = randn(1000)
+    @test 0.3 < StatsPlots.wand_bins(data) < 0.4
+end
+
 @testset "Distributions" begin
     @testset "univariate" begin
         @testset "discrete" begin