Dev

.github/workflows/ci.yml

@@ -29,7 +29,7 @@ jobs:
         with:
           version: ${{ matrix.version }}
           arch: ${{ matrix.arch }}
-      - uses: actions/cache@v1
+      - uses: actions/cache@v4
         env:
           cache-name: cache-artifacts
         with:

.gitignore

@@ -24,4 +24,7 @@ docs/site/
 Manifest.toml
 
 # script for testing
-script.jl
+script.jl
+
+.vscode/
+.DS_Store

Project.toml

@@ -1,7 +1,7 @@
 name = "QuantileMatching"
 uuid = "c8fc808f-91d0-4910-90d7-c7fb3b64f671"
 authors = ["Gabriel Gobeil <gabriel.gobeil@polymtl.ca>", "Jonathan Jalbert <jonathan.jalbert@polymtl.ca>"]
-version = "0.1.0"
+version = "0.2.0"
 
 [deps]
 CSV = "336ed68f-0bac-5ca0-87d4-7b16caf5d00b"

ReleaseNotes.md

@@ -0,0 +1,9 @@
+# Release Notes
+
+## 0.2.0
+
+- Refactored `match()` for scalar matching.  
+  The function can now be broadcast to match a vector.
+
+- Added an exception to parametric matching when the actual value to match is outside the support of the actual distribution.  
+  This situation can occur when the actual value lies outside the support of the actual distribution while the target distribution is unbounded. In such cases, the returned matched value is `Inf`.

docs/src/StationaryQuantileMatching.md

@@ -134,7 +134,7 @@ qmm = EmpiricalQuantileMatchingModel(y⁺, x⁺)
 
 Quantile matching of the non-zero simulated precipitations:
 ```@example stationary
-x̃⁺ = match(qmm, x⁺)
+x̃⁺ = match.(qmm, x⁺)
 
 println("") # hide
 ```
@@ -212,7 +212,7 @@ qmm = ParametricQuantileMatchingModel(fd_Y, fd_X)
 
 Quantile matching of the non-zero simulated precipitations:
 ```@example stationary
-x̃⁺ = match(qmm, x⁺)
+x̃⁺ = match.(qmm, x⁺)
 
 println("") # hide
 ```
@@ -304,7 +304,7 @@ qmm = ParametricQuantileMatchingModel(fd_Y, fd_X)
 
 Quantile matching of the non-zero simulated precipitations:
 ```@example stationary
-x̃⁺ = match(qmm, x⁺)
+x̃⁺ = match.(qmm, x⁺)
 
 println("") # hide
 ```
@@ -369,7 +369,7 @@ qmm = ParametricQuantileMatchingModel(fd_Y, fd_X)
 
 Quantile matching of the non-zero simulated precipitations:
 ```@example stationary
-x̃⁺ = match(qmm, x⁺)
+x̃⁺ = match.(qmm, x⁺)
 
 println("") # hide
 ```

src/AbstractQuantileMatchingModel.jl

@@ -9,6 +9,8 @@ abstract type AbstractQuantileMatchingModel end
 abstract type Stationary end
 abstract type NonStationary end
 
+Base.Broadcast.broadcastable(obj::AbstractQuantileMatchingModel) = Ref(obj)
+
 """
     match(qmm::AbstractQuantileMappingModel, x::Vector{<:Real})
 

src/AbstractQuantileMatchingModel/EmpiricalQuantileMatchingModel.jl

@@ -105,7 +105,7 @@ function showEmpiricalQuantileMatchingModel(io::IO, obj::EmpiricalQuantileMatchi
     println(io, prefix, "    ", "extrapolation: ",get_extrapolation(obj))
 end
 
-function match(eqmm::EmpiricalQuantileMatchingModel{Stationary}, x::Vector{<:Real})
+function match(eqmm::EmpiricalQuantileMatchingModel{Stationary}, x::Real)
 
     nbins = get_nbins(eqmm)
 
@@ -123,7 +123,7 @@ function match(eqmm::EmpiricalQuantileMatchingModel{Stationary}, x::Vector{<:Rea
     x̃ = itp(x)
 end
 
-function match(eqmm::EmpiricalQuantileMatchingModel{NonStationary}, x::Vector{<:Real})
+function match(eqmm::EmpiricalQuantileMatchingModel{NonStationary}, x::Real)
 
     nbins = get_nbins(eqmm)
 

src/AbstractQuantileMatchingModel/ParametricQuantileMatchingModel.jl

@@ -58,20 +58,43 @@ function showParametricQuantileMatchingModel(io::IO, obj::ParametricQuantileMatc
 end
 
 
-function match(pqm::ParametricQuantileMatchingModel{Stationary}, x::Vector{<:Real})
+"""
+    match(pqm::ParametricQuantileMatchingModel{Stationary}, x::Real)
+
+Parametric quantile matching of the actual value `x` according to the parametric quantile matching model `pqm`.
+
+### Note
+
+If `x` is outside the support of the actual distribution and the target distribution is unbounded, the function returns an infinite value.
+"""
+function match(pqm::ParametricQuantileMatchingModel{Stationary}, x::Real)
 
     targetdist = get_targetdist(pqm)
     actualdist = get_actualdist(pqm)
 
-    p = cdf.(actualdist, x)
-
-    x̃ = quantile.(targetdist, p)
+    p = cdf(actualdist, x)
 
+    if p ≈ 0.
+        if isfinite(minimum(targetdist))
+            x̃ = minimum(targetdist)
+        else
+            x̃ = -Inf
+        end
+    elseif p ≈ 1.
+        if isfinite(maximum(targetdist))
+            x̃ = maximum(targetdist)
+        else
+            x̃ = Inf
+        end
+    else
+        x̃ = quantile.(targetdist, p)
+    end
+
     return x̃
 
 end
 
-function match(nspqm::ParametricQuantileMatchingModel{NonStationary}, x::Vector{<:Real})
+function match(nspqm::ParametricQuantileMatchingModel{NonStationary}, x::Real)
 
     targetdist = get_targetdist(nspqm)
     actualdist = get_actualdist(nspqm)

src/functions.jl

@@ -51,7 +51,7 @@ function eqm(y::Vector{<:Real}, x::Vector{<:Real})
     qmm = EmpiricalQuantileMatchingModel(y⁺, x⁺)
 
     # Quantile matching of non-zero values
-    x̃⁺ = match(qmm, x⁺)
+    x̃⁺ = match.(qmm, x⁺)
 
     # Replace the non-zero values in the frequency adjusted series.
     x̃[x̃ .> 0] = x̃⁺
@@ -89,7 +89,7 @@ function pqm(pd::Type{<:ContinuousUnivariateDistribution}, y::AbstractVector{<:R
     qmm = ParametricQuantileMatchingModel(fd_Y, fd_X)
 
     # Quantile matching of non-zero values
-    x̃⁺ = match(qmm, x⁺)
+    x̃⁺ = match.(qmm, x⁺)
 
     # Replace the non-zero values in the frequency adjusted series. 
     x̃[x̃ .> 0] = x̃⁺

test/AbstractQuantileMatchingModel/EmpiricalQuantileMatchingModel_test.jl

@@ -65,15 +65,15 @@ end
 
         qmm = EmpiricalQuantileMatchingModel(targetsample, actualsample)
 
-        @test match(qmm, [2.])[] ≈ 0.48401845735265503
+        @test match(qmm, 2.) ≈ 0.48401845735265503
 
     end
 
     @testset "non-stationary" begin
 
         qmm = EmpiricalQuantileMatchingModel(targetsample, actualsample, projsample)
 
-        @test match(qmm, [2.])[] ≈ 0.8349460253709341
+        @test match(qmm, 2.) ≈ 0.8349460253709341
 
     end
 

test/AbstractQuantileMatchingModel/ParametricQuantileMatchingModel_test.jl

@@ -52,33 +52,56 @@ end
 
 @testset "match(::ParametricQuantileMatchingModel)" begin
 
-    targetdist = Exponential(.5)
-    actualdist = Exponential(2)
-    projdist = Exponential(3)
-
     @testset "stationary" begin
+
+        targetdist = Exponential(.5)
+        actualdist = Exponential(2)
 
         qmm = ParametricQuantileMatchingModel(targetdist, actualdist)
 
-        x = [4]
-        x̃ = rate(actualdist)/rate(targetdist)*[4]
+        x = 4
+        x̃ = rate(actualdist)/rate(targetdist)*4
 
         @test match(qmm, x) ≈ x̃
 
     end
 
     @testset "non-stationary" begin
+
+        targetdist = Exponential(.5)
+        actualdist = Exponential(2)
+        projdist = Exponential(3)
 
         qmm = ParametricQuantileMatchingModel(targetdist, actualdist, projdist)
 
         λ = rate(targetdist)*rate(projdist)/rate(actualdist)
 
-        x = [4.]
+        x = 4.
         x̃ = quantile(Exponential(1/λ), cdf(projdist, x))
 
         @test match(qmm, x) ≈ x̃
 
     end
+
+    @testset "outside the support" begin
+
+        targetdist = Exponential(1)
+        actualdist = Beta(1,1)
+
+        qmm = ParametricQuantileMatchingModel(targetdist, actualdist)
+
+        @test isfinite(match(qmm, -1.))
+        @test !isfinite(match(qmm, 1.2))
+
+        targetdist = Beta(2,2)
+        qmm = ParametricQuantileMatchingModel(targetdist, actualdist)
+        @test isfinite(match(qmm, 1.2))
+
+        targetdist = Normal(0,1)
+        qmm = ParametricQuantileMatchingModel(targetdist, actualdist)
+        @test !isfinite(match(qmm, -1.))
+
+    end
 
 end
 

test/functions_test.jl

@@ -24,7 +24,7 @@
         actualsample = [0.21713363327549823, 1.502901744984021, 2.046092118287037, 2.2249912757196464, 2.6629189261859207]
 
         qmm = EmpiricalQuantileMatchingModel(targetsample, actualsample)
-        x̃ = match(qmm, actualsample)
+        x̃ = match.(qmm, actualsample)
 
         @test all(QuantileMatching.eqm(targetsample, actualsample) .≈ x̃)
 
@@ -40,7 +40,7 @@
 
         qmm = ParametricQuantileMatchingModel(fd_Y, fd_X)
 
-        x̃ = match(qmm, actualsample)
+        x̃ = match.(qmm, actualsample)
 
         @test all(pqm(Gamma, targetsample, actualsample) .≈ x̃)