replace ols(X,y) with X\y in other methods

Author: jbytecode

src/hs93.jl

@@ -6,7 +6,7 @@ export hs93
 
 import ..Basis:
     RegressionSetting, @extractRegressionSetting, designMatrix, responseVector, applyColumns
-import ..OrdinaryLeastSquares: ols, predict, residuals, coef
+
 import ..Diagnostics: dffits
 
 import Distributions: TDist, quantile
@@ -116,8 +116,7 @@ function hs93basicsubset(
     s = length(initialindices)
     indices = initialindices
     for i in range(s + 1, stop = h)
-        olsreg = ols(X[indices, :], y[indices])
-        betas = coef(olsreg)
+        betas = X[indices, :] \ y[indices]
         d = zeros(Float64, n)
         XM = X[indices, :]
         for j = 1:n
@@ -204,9 +203,8 @@ function hs93(
     betas = []
 
     while s < n
-        olsreg = ols(X[indices, :], y[indices])
-        betas = coef(olsreg)
-        resids = residuals(olsreg)
+        betas = X[indices, :] \ y[indices]
+        resids = y[indices] - X[indices,:] * betas
         sigma = sqrt(sum(resids .^ 2.0) / (length(resids) - p))
         d = zeros(Float64, n)
         XM = X[indices, :]
@@ -237,9 +235,7 @@ function hs93(
 
         outlierset = filter(x -> abs(d[x]) > abs(tcalc), 1:n)
         inlierset = setdiff(1:n, outlierset)
-        cleanols = ols(X[inlierset, :], y[inlierset])
-        cleanbetas = coef(cleanols)
-
+        cleanbetas = X[inlierset, :] \ y[inlierset]
         if abs(d[orderingd][s+1]) > abs(tcalc)
             result = Dict(
                 "d" => d,

src/satman2013.jl

@@ -7,7 +7,7 @@ export satman2013
 import ..Basis:
     RegressionSetting, @extractRegressionSetting, designMatrix, responseVector, applyColumns
 import ..LTS: iterateCSteps
-import ..OrdinaryLeastSquares: ols, coef, wls, residuals
+import ..OrdinaryLeastSquares: coef, wls, residuals
 import ..Diagnostics: mahalanobisSquaredMatrix
 import Distributions: median
 import LinearAlgebra: diag
@@ -106,9 +106,8 @@ function satman2013(X::AbstractMatrix{Float64}, y::AbstractVector{Float64})
     _, bestset = iterateCSteps(X, y, best_h_indices, h)
 
     # Estimate the final regression parameters
-    olsreg = ols(X[bestset, :], y[bestset])
-    betas = coef(olsreg)
-    resids = y .- (X * betas)
+    betas = X[bestset, :] \ y[bestset]
+    resids = y .- X * betas
     med_res = median(resids)
     standardized_resids = (resids .- med_res) / median(abs.(resids .- med_res))
 

src/theilsen.jl

@@ -46,9 +46,9 @@ function theilsen(X::AbstractMatrix{Float64}, y::AbstractVector{Float64}, m::Int
 
     for i in 1:nsamples
         luckyindices = sample(1:n, m, replace = false)
-        olsresult = ols(X[luckyindices, :], y[luckyindices])
-        betas = coef(olsresult)
-        allbetas[i, :] = betas
+        #olsresult = ols(X[luckyindices, :], y[luckyindices])
+        #betas = coef(olsresult)
+        allbetas[i, :] = X[luckyindices, :] \ y[luckyindices]
     end 
 
     multimed = multivariatemedian(allbetas)