Aligned names of ess functions

Author: Closed-Limelike-Curves

.lh/src/ESS.jl.json

@@ -3,7 +3,7 @@
     "activeCommit": 0,
     "commits": [
         {
-            "activePatchIndex": 4,
+            "activePatchIndex": 11,
             "patches": [
                 {
                     "date": 1626309971342,
@@ -24,6 +24,34 @@
                 {
                     "date": 1626314567863,
                     "content": "Index: \n===================================================================\n--- \n+++ \n@@ -1,8 +1,7 @@\n using FFTW\n using MCMCDiagnosticTools\n using LoopVectorization\n-using TensorOperations\n using Tullio\n \n export relative_eff, psis_n_eff\n \n"
+                },
+                {
+                    "date": 1626454088792,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -2,14 +2,15 @@\n using MCMCDiagnosticTools\n using LoopVectorization\n using Tullio\n \n-export relative_eff, psis_n_eff\n+export relative_eff, psis_ess\n \n \"\"\"\n     relative_eff(sample::AbstractArray{AbstractFloat, 3}; method=FFTESSMethod())\n \n-Compute the MCMC effective sample size divided by the nominal sample size.\n+Calculate the relative efficiency of an MCMC chain, i.e. the effective sample size divided\n+by the nominal sample size.\n \"\"\"\n function relative_eff(\n     sample::AbstractArray{T,3}; method=FFTESSMethod()\n ) where {T<:AbstractFloat}\n@@ -22,28 +23,37 @@\n     return r_eff\n end\n \n \"\"\"\n-    function psis_n_eff(\n+    function psis_ess(\n         weights::AbstractVector{T},\n         r_eff::AbstractVector{T}\n     ) -> AbstractVector{T}\n+\n+Calculate the (approximate) effective sample size of a PSIS sample, using the correction in\n+Vehtari et al. 2019.\n+\n+# Arguments\n+\n+- `weights`: A set of importance sampling weights derived from PSIS.\n+- `r_eff`: The relative efficiency of the MCMC chains from which PSIS samples were derived.\n+See `?relative_eff` to calculate `r_eff`.\n \"\"\"\n-function psis_n_eff(\n+function psis_ess(\n     weights::AbstractVector{T}, r_eff::AbstractVector{T}\n ) where {T<:AbstractFloat}\n     @tullio sum_of_squares := weights[x]^2\n     return r_eff ./ sum_of_squares\n end\n \n-function psis_n_eff(\n+function psis_ess(\n     weights::AbstractMatrix{T}, r_eff::AbstractVector{T}\n ) where {T<:AbstractFloat}\n     @tullio sum_of_squares[x] := weights[x, y]^2\n     return @tturbo r_eff ./ sum_of_squares\n end\n \n-function psis_n_eff(weights::AbstractArray{T}) where {T<:AbstractFloat}\n+function psis_ess(weights::AbstractArray{T}) where {T<:AbstractFloat}\n     @warn \"PSIS ESS not adjusted based on MCMC ESS. MCSE and ESS estimates \" *\n           \"will be overoptimistic if samples are autocorrelated.\"\n-    return psis_n_eff(weights, ones(size(weights)))\n+    return psis_ess(weights, ones(size(weights)))\n end\n"
+                },
+                {
+                    "date": 1626454315198,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -56,4 +56,8 @@\n     @warn \"PSIS ESS not adjusted based on MCMC ESS. MCSE and ESS estimates \" *\n           \"will be overoptimistic if samples are autocorrelated.\"\n     return psis_ess(weights, ones(size(weights)))\n end\n+\n+function psis_n_eff(args...) \n+    psis_ess\n+end\n\\ No newline at end of file\n"
+                },
+                {
+                    "date": 1626454380380,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -2,9 +2,9 @@\n using MCMCDiagnosticTools\n using LoopVectorization\n using Tullio\n \n-export relative_eff, psis_ess\n+export relative_eff, psis_ess, psis_n_eff\n \n \"\"\"\n     relative_eff(sample::AbstractArray{AbstractFloat, 3}; method=FFTESSMethod())\n \n"
+                },
+                {
+                    "date": 1626454430031,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -51,13 +51,13 @@\n     @tullio sum_of_squares[x] := weights[x, y]^2\n     return @tturbo r_eff ./ sum_of_squares\n end\n \n-function psis_ess(weights::AbstractArray{T}) where {T<:AbstractFloat}\n+function psis_ess(weights::AbstractMatrix{T}) where {T<:AbstractFloat}\n     @warn \"PSIS ESS not adjusted based on MCMC ESS. MCSE and ESS estimates \" *\n           \"will be overoptimistic if samples are autocorrelated.\"\n     return psis_ess(weights, ones(size(weights)))\n end\n \n-function psis_n_eff(args...) \n-    psis_ess\n+function psis_n_eff(args...; kwargs...) \n+    return psis_ess(args...; kwargs...)\n end\n\\ No newline at end of file\n"
+                },
+                {
+                    "date": 1626454496697,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -57,7 +57,7 @@\n           \"will be overoptimistic if samples are autocorrelated.\"\n     return psis_ess(weights, ones(size(weights)))\n end\n \n-function psis_n_eff(args...; kwargs...) \n+function psis_n_eff(args...; kwargs...)  # Alias for compatibility with R version\n     return psis_ess(args...; kwargs...)\n end\n\\ No newline at end of file\n"
+                },
+                {
+                    "date": 1626454514402,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -2,9 +2,9 @@\n using MCMCDiagnosticTools\n using LoopVectorization\n using Tullio\n \n-export relative_eff, psis_ess, psis_n_eff\n+export relative_eff, psis_ess\n \n \"\"\"\n     relative_eff(sample::AbstractArray{AbstractFloat, 3}; method=FFTESSMethod())\n \n@@ -56,8 +56,4 @@\n     @warn \"PSIS ESS not adjusted based on MCMC ESS. MCSE and ESS estimates \" *\n           \"will be overoptimistic if samples are autocorrelated.\"\n     return psis_ess(weights, ones(size(weights)))\n end\n-\n-function psis_n_eff(args...; kwargs...)  # Alias for compatibility with R version\n-    return psis_ess(args...; kwargs...)\n-end\n\\ No newline at end of file\n"
+                },
+                {
+                    "date": 1626454525488,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -2,9 +2,9 @@\n using MCMCDiagnosticTools\n using LoopVectorization\n using Tullio\n \n-export relative_eff, psis_ess\n+export relative_eff, psis_ess, psis_n_eff\n \n \"\"\"\n     relative_eff(sample::AbstractArray{AbstractFloat, 3}; method=FFTESSMethod())\n \n@@ -56,4 +56,8 @@\n     @warn \"PSIS ESS not adjusted based on MCMC ESS. MCSE and ESS estimates \" *\n           \"will be overoptimistic if samples are autocorrelated.\"\n     return psis_ess(weights, ones(size(weights)))\n end\n+\n+function psis_n_eff(args...; kwargs...)  # Alias for compatibility with R version\n+    return psis_ess(args...; kwargs...)\n+end\n\\ No newline at end of file\n"
                 }
             ],
             "date": 1626309971342,

.lh/src/ImportanceSampling.jl.json

@@ -3,7 +3,7 @@
     "activeCommit": 0,
     "commits": [
         {
-            "activePatchIndex": 33,
+            "activePatchIndex": 36,
             "patches": [
                 {
                     "date": 1626309654641,
@@ -140,6 +140,18 @@
                 {
                     "date": 1626452759502,
                     "content": "Index: \n===================================================================\n--- \n+++ \n@@ -1,6 +1,5 @@\n using LoopVectorization\n-using Polyester\n using Tullio\n \n \n const LIKELY_ERROR_CAUSES = \"\"\"\n"
+                },
+                {
+                    "date": 1626453064337,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -105,9 +105,9 @@\n end\n \n \n \"\"\"\n-    do_psis_i!(is_ratios::AbstractVector{AbstractFloat}, tail_length::Integer)::T\n+    do_psis_i!(is_ratios::AbstractVector{AbstractFloat}, tail_length::Integer) -> T\n \n Do PSIS on a single vector, smoothing its tail values.\n \n # Arguments\n"
+                },
+                {
+                    "date": 1626453089719,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -242,8 +242,9 @@\n         return r_eff\n     end\n end\n \n+\n \"\"\"\n Make sure all inputs to `psis` are valid.\n \"\"\"\n function check_input_validity_psis(\n@@ -266,8 +267,9 @@\n     end\n     return nothing\n end\n \n+\n \"\"\"\n Check the tail to make sure a GPD fit is possible.\n \"\"\"\n function check_tail(tail::AbstractVector{T}) where {T<:AbstractFloat}\n"
+                },
+                {
+                    "date": 1626454092733,
+                    "content": "Index: \n===================================================================\n--- \n+++ \n@@ -71,9 +71,9 @@\n     end\n     \n     @tullio norm_const[i] := weights[i, j]\n     weights .= weights ./ norm_const\n-    ess = psis_n_eff(weights, r_eff)\n+    ess = psis_ess(weights, r_eff)\n \n     weights = reshape(weights, dims)\n \n     if log_weights\n"
                 }
             ],
             "date": 1626309654640,

src/ESS.jl

@@ -3,12 +3,13 @@ using MCMCDiagnosticTools
 using LoopVectorization
 using Tullio
 
-export relative_eff, psis_n_eff
+export relative_eff, psis_ess, psis_n_eff
 
 """
     relative_eff(sample::AbstractArray{AbstractFloat, 3}; method=FFTESSMethod())
 
-Compute the MCMC effective sample size divided by the nominal sample size.
+Calculate the relative efficiency of an MCMC chain, i.e. the effective sample size divided
+by the nominal sample size.
 """
 function relative_eff(
     sample::AbstractArray{T,3}; method=FFTESSMethod()
@@ -23,27 +24,40 @@ function relative_eff(
 end
 
 """
-    function psis_n_eff(
+    function psis_ess(
         weights::AbstractVector{T},
         r_eff::AbstractVector{T}
     ) -> AbstractVector{T}
+
+Calculate the (approximate) effective sample size of a PSIS sample, using the correction in
+Vehtari et al. 2019.
+
+# Arguments
+
+- `weights`: A set of importance sampling weights derived from PSIS.
+- `r_eff`: The relative efficiency of the MCMC chains from which PSIS samples were derived.
+See `?relative_eff` to calculate `r_eff`.
 """
-function psis_n_eff(
+function psis_ess(
     weights::AbstractVector{T}, r_eff::AbstractVector{T}
 ) where {T<:AbstractFloat}
     @tullio sum_of_squares := weights[x]^2
     return r_eff ./ sum_of_squares
 end
 
-function psis_n_eff(
+function psis_ess(
     weights::AbstractMatrix{T}, r_eff::AbstractVector{T}
 ) where {T<:AbstractFloat}
     @tullio sum_of_squares[x] := weights[x, y]^2
     return @tturbo r_eff ./ sum_of_squares
 end
 
-function psis_n_eff(weights::AbstractArray{T}) where {T<:AbstractFloat}
+function psis_ess(weights::AbstractMatrix{T}) where {T<:AbstractFloat}
     @warn "PSIS ESS not adjusted based on MCMC ESS. MCSE and ESS estimates " *
           "will be overoptimistic if samples are autocorrelated."
-    return psis_n_eff(weights, ones(size(weights)))
+    return psis_ess(weights, ones(size(weights)))
 end
+
+function psis_n_eff(args...; kwargs...)  # Alias for compatibility with R version
+    return psis_ess(args...; kwargs...)
+end

src/ImportanceSampling.jl

@@ -72,7 +72,7 @@ function psis(
 
     @tullio norm_const[i] := weights[i, j]
     weights .= weights ./ norm_const
-    ess = psis_n_eff(weights, r_eff)
+    ess = psis_ess(weights, r_eff)
 
     weights = reshape(weights, dims)
 
@@ -106,7 +106,7 @@ end
 
 
 """
-    do_psis_i!(is_ratios::AbstractVector{AbstractFloat}, tail_length::Integer)::T
+    do_psis_i!(is_ratios::AbstractVector{AbstractFloat}, tail_length::Integer) -> T
 
 Do PSIS on a single vector, smoothing its tail values.
 
@@ -243,6 +243,7 @@ function _generate_r_eff(weights, dims, r_eff, source)
     end
 end
 
+
 """
 Make sure all inputs to `psis` are valid.
 """
@@ -267,6 +268,7 @@ function check_input_validity_psis(
     return nothing
 end
 
+
 """
 Check the tail to make sure a GPD fit is possible.
 """