Add and test simple nanargmin and nanargmax

Project.toml

@@ -1,7 +1,7 @@
 name = "NaNStatistics"
 uuid = "b946abbf-3ea7-4610-9019-9858bfdeaf2d"
 authors = ["C. Brenhin Keller"]
-version = "0.6.23"
+version = "0.6.24"
 
 [deps]
 IfElse = "615f187c-cbe4-4ef1-ba3b-2fcf58d6d173"

src/ArrayStats/ArrayStats.jl

@@ -141,7 +141,7 @@
     ```julia
     nanadd!(A, B)
     ```
-    Add the non-NaN elements of `B` to `A`, treating NaNs as zeros
+    Add the non-NaN elements of `B` to `A`, treating `NaN`s as zeros
     """
     function nanadd!(A::Array, B::AbstractArray)
         @inbounds @simd for i ∈ eachindex(A)
@@ -181,7 +181,7 @@
     ```julia
     nanmaximum(A; dims)
     ```
-    Find the largest non-NaN value of an indexable collection `A`, optionally
+    As `maximum` but ignoring `NaN`s: Find the largest non-`NaN` value of an indexable collection `A`, optionally
     along a dimension specified by `dims`.
 
     Also supports the `dim` keyword, which behaves identically to `dims`, but
@@ -197,13 +197,50 @@
     _nanmaximum(A::Array{<:Number}, ::Colon) = vreduce(nanmax, A)
     export nanmaximum
 
+    """
+    ```julia
+    nanargmin(A)
+    ```
+    As `argmin` but ignoring `NaN`s: Find the index of the smallest non-`NaN`
+    value (if any) of an indexable collection `A`
+    """
+    function nanargmin(x)
+        imin = firstindex(x)
+        @inbounds for i in eachindex(x)
+            xᵢ = x[i]
+            if xᵢ==xᵢ && !(xᵢ > x[imin])
+                imin = i
+            end
+        end
+        return imin
+    end
+    export nanargmin
+
+    """
+    ```julia
+    nanargmax(A)
+    ```
+    As `argmax` but ignoring `NaN`s: Find the index of the largest non-`NaN`
+    value (if any) of an indexable collection `A`
+    """
+    function nanargmax(x)
+        imax = firstindex(x)
+        @inbounds for i in eachindex(x)
+            xᵢ = x[i]
+            if xᵢ==xᵢ && !(xᵢ < x[imax])
+                imax = i
+            end
+        end
+        return imax
+    end
+    export nanargmax
 
     """
     ```julia
     nanextrema(A; dims)
     ```
     Find the extrema (maximum & minimum) of an indexable collection `A`,
-    ignoring NaNs, optionally along a dimension specified by `dims`.
+    ignoring `NaN`s, optionally along a dimension specified by `dims`.
 
     Also supports the `dim` keyword, which behaves identically to `dims`, but
     also drops any singleton dimensions that have been reduced over (as is the
@@ -221,7 +258,7 @@
     nanrange(A; dims)
     ```
     Calculate the range (maximum - minimum) of an indexable collection `A`,
-    ignoring NaNs, optionally along a dimension specified by `dims`.
+    ignoring `NaN`s, optionally along a dimension specified by `dims`.
 
     Also supports the `dim` keyword, which behaves identically to `dims`, but
     also drops any singleton dimensions that have been reduced over (as is the
@@ -235,7 +272,7 @@
     ```julia
     nanmean(A, W; dims)
     ```
-    Ignoring NaNs, calculate the weighted mean of an indexable
+    Ignoring `NaN`s, calculate the weighted mean of an indexable
     collection `A`, optionally along dimensions specified by `dims`.
 
     Also supports the `dim` keyword, which behaves identically to `dims`, but
@@ -295,7 +332,7 @@
     ```julia
     nanstd(A, W; dims)
     ```
-    Calculate the weighted standard deviation, ignoring NaNs, of an
+    Calculate the weighted standard deviation, ignoring `NaN`s, of an
     indexable collection `A`, optionally along a dimension specified by `dims`.
 
     Also supports the `dim` keyword, which behaves identically to `dims`, but
@@ -374,7 +411,7 @@
     ```julia
     nanmad(A; dims)
     ```
-    Median absolute deviation from the median, ignoring NaNs, of an indexable
+    Median absolute deviation from the median, ignoring `NaN`s, of an indexable
     collection `A`, optionally along a dimension specified by `dims`.
     Note that for a Normal distribution, sigma = 1.4826 * MAD.
 
@@ -403,7 +440,7 @@
     ```julia
     nanaad(A; dims)
     ```
-    Mean (average) absolute deviation from the mean, ignoring NaNs, of an
+    Mean (average) absolute deviation from the mean, ignoring `NaN`s, of an
     indexable collection `A`, optionally along a dimension specified by `dims`.
     Note that for a Normal distribution, sigma = 1.253 * AAD.
 

src/ArrayStats/nancumsum.jl

@@ -2,7 +2,7 @@
 ```julia
 nancumsum(A; dims)
 ```
-Calculate the sum of an indexable collection `A`, ignoring NaNs, optionally
+Calculate the sum of an indexable collection `A`, ignoring `NaN`s, optionally
 along dimensions specified by `dims`.
 
 ## Examples

src/ArrayStats/nansum.jl

@@ -2,7 +2,7 @@
 ```julia
 nansum(A; dims)
 ```
-Calculate the sum of an indexable collection `A`, ignoring NaNs, optionally
+Calculate the sum of an indexable collection `A`, ignoring `NaN`s, optionally
 along dimensions specified by `dims`.
 
 Also supports the `dim` keyword, which behaves identically to `dims`, but

src/Binning.jl

@@ -4,7 +4,7 @@
     ```julia
     nanbinmean!(MU, [N], x, y, xedges::AbstractRange)
     ```
-    Ignoring NaNs, fill the array `MU` with the means (and optionally `N` with
+    Ignoring `NaN`s, fill the array `MU` with the means (and optionally `N` with
     the counts) of non-NAN `y` values that fall into each of `length(xedges)-1`
     equally spaced bins along the `x` axis with bin edges specified by `xedges`.
 
@@ -97,7 +97,7 @@
     ```julia
     nanbinmean!(MU, N, x, y, z, xedges::AbstractRange, yedges::AbstractRange)
     ```
-    Ignoring NaNs, fill the matrix `MU` with the means and `N` with
+    Ignoring `NaN`s, fill the matrix `MU` with the means and `N` with
     the counts of non-NAN `z` values that fall into a 2D grid of x and y bins
     defined by `xedges` and `yedges`. The independent variables `x` and `y`,
     as well as the dependent variable `z`, are all expected as 1D vectors (any
@@ -150,7 +150,7 @@
     ```julia
     nanbinmean!(MU, W, x, y, w, xedges::AbstractRange)
     ```
-    Ignoring NaNs, fill the array `MU` with the weighted means (and `W` with
+    Ignoring `NaN`s, fill the array `MU` with the weighted means (and `W` with
     the sum of weights) of non-NAN `y` values that fall into each of
     `length(xedges)-1` equally spaced bins along the `x` axis with bin edges
     specified by `xedges`.
@@ -308,7 +308,7 @@
     ```julia
     nanbinmean(x, y, xedges::AbstractRange)
     ```
-    Ignoring NaNs, calculate the mean of `y` values that fall into each of
+    Ignoring `NaN`s, calculate the mean of `y` values that fall into each of
     `length(xedges)-1` equally spaced bins along the `x` axis with bin edges
     specified by `xedges`.
 
@@ -346,7 +346,7 @@
     ```julia
     nanbinmean(x, y, z, xedges, yedges)
     ```
-    Ignoring NaNs, calculate the mean of `z` values that fall into a 2D grid of
+    Ignoring `NaN`s, calculate the mean of `z` values that fall into a 2D grid of
     x and y bins with bin edges defined by `xedges` and `yedges`. The independent
     variables `x` and `y`, as well as the dependent variable `z`, are all expected
     as 1D vectors (any subtype of AbstractVector).
@@ -381,7 +381,7 @@
     ```julia
     nanbinmean(x, y, xedges::AbstractRange)
     ```
-    Ignoring NaNs, calculate the weighted mean of `y` values that
+    Ignoring `NaN`s, calculate the weighted mean of `y` values that
     fall into each of `length(xedges)-1` equally spaced bins along the `x`
     axis with bin edges specified by `xedges`.
 
@@ -402,7 +402,7 @@
     ```julia
     nanbinmedian(x, y, xedges::AbstractRange)
     ```
-    Calculate the median, ignoring NaNs, of y values that fall into each of
+    Calculate the median, ignoring `NaN`s, of y values that fall into each of
     `length(xedges)-1` equally spaced bins along the `x` axis with bin edges
     specified by `xedges`.
 

src/Histograms.jl

@@ -3,7 +3,7 @@
 ```julia
 histcounts(x, xedges::AbstractRange; T=Int64)::Vector{T}
 ```
-A 1D histogram, ignoring NaNs: calculate the number of `x` values that fall into
+A 1D histogram, ignoring `NaN`s: calculate the number of `x` values that fall into
 each of `length(xedges)-1` equally spaced bins along the `x` axis with bin edges
 specified by `xedges`.
 
@@ -40,7 +40,7 @@ histcounts(x, xmin::Number, xmax::Number, nbins::Integer; T=Int64) = histcounts(
 ```julia
 histcountindices(x, xedges::AbstractRange; T=Int64)::Vector{T}
 ```
-A 1D histogram, ignoring NaNs; as `histcounts` but also returning a vector of
+A 1D histogram, ignoring `NaN`s; as `histcounts` but also returning a vector of
 the bin index of each `x` value.
 
 ## Examples
@@ -65,7 +65,7 @@ export histcountindices
 ```julia
 histcounts(x, y, xedges::AbstractRange, yedges::AbstractRange; T=Int64)::Matrix{T}
 ```
-A 2D histogram, ignoring NaNs: calculate the number of `x, y` pairs that fall into
+A 2D histogram, ignoring `NaN`s: calculate the number of `x, y` pairs that fall into
 each square of a 2D grid of equally-spaced square bins with edges specified by
 `xedges` and `yedges`.
 

src/precompile.jl

@@ -29,6 +29,8 @@
                 nanvar(A)
                 nanminimum(A)
                 nanmaximum(A)
+                nanargmin(A)
+                nanargmax(A)
                 # nanmedian(A)
                 # nanpctile(A, 50)