LMS returns betas rather than coefs

Author: jbytecode

CHANGELOG.md

@@ -1,5 +1,6 @@
-# v0.9.6 (Upcoming Release)
+# v0.10.0 (Upcoming Release)
 
+- LMS returns betas rather than coefs
 
 
 # v0.9.5 

src/lms.jl

@@ -94,7 +94,7 @@ function lms(X::Array{Float64,2}, y::Array{Float64,1}; iters = nothing, crit = 2
     s = 1.4826 * sqrt((1.0 + (5.0 / (n - p))) * bestobjective)
     standardizedres = bestres / s
     d = Dict()
-    d["coef"] = bestparamaters
+    d["betas"] = bestparamaters
     d["objective"] = bestobjective
     d["S"] = s
     d["stdres"] = standardizedres

test.jl

@@ -1,21 +1,39 @@
 using LinRegOutliers
 
-
+import LinRegOutliers.OrdinaryLeastSquares: ols, wls, coef, residuals
+import Distributions: median
 
 #sett = createRegressionSetting(@formula(calls ~ year), phones)
 #sett = createRegressionSetting(@formula(y ~ x1 + x2 + x3), hbk)
 
-n = 1000
-p = 25
-betas = [5.0 for i in 1:p]
-x = rand(n, p)
-e = randn(n)
-y = x * betas + e
-range = collect(701:1000)
-no = length(range)
-
-y[range] .= rand(no)
-for i in 1:25
-    x[range, i] .= rand(no)
+function generatedata(
+    n::Int,
+    p::Int,
+    contamination::Float64,
+    direction::Symbol,
+)::Tuple{Array{Float64,2},Array{Float64,1}}
+    if !(direction in [:y, :x])
+        error("- Direction must be :x or :y")
+    end 
+    on = ones(Float64, n)
+    pxvars = p - 1 
+    xvars = randn(n, pxvars)
+    b = [5.0 for i = 1:pxvars]
+    y = 5.0 .+ xvars * b + randn(n)
+    totalcontamination = round(Int, contamination * n)
+    if direction == :y
+        y[1:totalcontamination] .= maximum(y) .+ abs.(rand(totalcontamination) * 5.0)
+    elseif direction == :x
+        xvars[1:totalcontamination, :] .=
+            maximum(xvars) .+ abs.(rand(totalcontamination) * 5.0)
+    end 
+    return (hcat(on, xvars), y)
 end
-result = lts(x, y)
+
+
+
+sett = createRegressionSetting(@formula(calls ~ year), phones)
+x = designMatrix(sett)
+y = responseVector(sett)
+x, y = generatedata(1000, 25, 0.30, :x)
+result = gwcga(x, y)