Implement VARHAC

Project.toml

@@ -9,7 +9,7 @@ GroupedArrays = "6407cd72-fade-4a84-8a1e-56e431fc1533"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
-
+NaNStatistics = "b946abbf-3ea7-4610-9019-9858bfdeaf2d"
 [weakdeps]
 GLM = "38e38edf-8417-5370-95a0-9cbb8c7f171a"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
@@ -23,6 +23,7 @@ Combinatorics = "1.0"
 GLM = "1.9"
 GroupedArrays = "0.3"
 julia = "1.9"
+NaNStatistics = "0.6"
 SparseArrays = "1.0"
 Statistics = "1.0"
 StatsBase = "0.34"

src/CR.jl

@@ -10,6 +10,18 @@ function clusterize(X::Matrix, g::GroupedArray)
     return Symmetric(X2' * X2)
 end
 
+function clusterize_mean(X::Matrix, g::GroupedArray)
+    X2 = zeros(eltype(X), g.ngroups, size(X, 2))
+    idx = 0
+    for j ∈ axes(X, 2)
+        idx += 1
+        @inbounds @simd for i ∈ axes(X, 1)
+            X2[g.groups[i], idx] += X[i, j]
+        end
+    end
+    return X2
+end
+
 function avar(k::T, X::Union{Matrix{F},Vector{F}};
               kwargs...) where {T<:Union{CR0,CR1,CR2,CR3},F<:AbstractFloat}
     f = k.g

src/CovarianceMatrices.jl

@@ -14,10 +14,10 @@ using GroupedArrays
 using LinearAlgebra
 using Statistics
 using StatsBase
+using NaNStatistics
 using Base.Threads
 
 include("types.jl")
-include("aVar.jl")
 include("HAC.jl")
 include("CR.jl")
 include("HR.jl")
@@ -26,6 +26,8 @@ include("demeaner.jl")
 include("EWC.jl")
 include("api.jl")
 include("smoothing.jl")
+include("VARHAC.jl")
+include("aVar.jl")
 #include("VARHAC.jl")
 export Andrews,
        NeweyWest,