implement shifloNormal_mode

docs/src/logitnormal.md

@@ -62,8 +62,12 @@ p # hide
 A shifted and scaled version of this distribution can be
 used as a moother alternative to the Bounded uniform distribution.
 
+When no information on mode or peakedness is required, prefer the option
+without mode, because those does not require optimization.
+
 ```@docs
 shifloNormal
+shifloNormal_mode
 ```
 
 

docs/src/partype.md

@@ -22,7 +22,7 @@ true
 
 If the parametric type is omitted, default Float64 is assumed, or inferred from
 other parameters of the fitting function.
-Since quantiles and median are rather sample-like measures than paraemter-like
+Since quantiles and median are rather sample-like measures than parameter-like
 measures, they do not influence the inferred parameter type.
 
 

src/DistributionFits.jl

@@ -35,7 +35,7 @@ export
 export AbstractΣstar, Σstar, σstar, ScaledLogNormal
 
 # LogitNormal  
-export fit_mode_flat, shifloNormal
+export fit_mode_flat, shifloNormal, shifloNormal_mode
 
 # dependency inversion: need to define DistributionFits.optimize by user
 export AbstractDistributionFitOptimizer, optimize

src/univariate/continuous/logitnormal.jl

@@ -113,13 +113,13 @@ true
 """
 function fit_mode_flat(::Type{LogitNormal},
     mode::T,
-    ::Val{nTry} = Val(40);
+    ntry::Val{nTry} = Val(40);
     peakedness = 1) where {T <: Real, nTry}
-    fit_mode_flat(LogitNormal{T}, mode, Val(nTry); peakedness)
+    fit_mode_flat(LogitNormal{T}, mode, ntry; peakedness)
 end
 function fit_mode_flat(::Type{LogitNormal{T}},
     mode,
-    ::Val{nTry} = Val(40);
+    ntry::Val{nTry} = Val(40);
     peakedness = 1) where {T <: Real, nTry}
     mode == 0.5 && return (LogitNormal{T}(zero(T), sqrt(2) / peakedness))
     is_right = mode > 0.5
@@ -158,3 +158,22 @@ function shifloNormal(lower, upper)
     #dln * (upper-lower) + lower
     Distributions.AffineDistribution{T}(lower, (upper - lower), dln)
 end
+
+"""
+    shifloNormal_mode(lower,upper,mode; peakedness=1.1)
+
+Shifted LogitNormal where mode and peakedness are specified.    
+Note, that this requires optimization, while shifloNormal without mode
+is a closed form solution.
+"""
+function shifloNormal_mode(lower, upper, mode,
+    ntry::Val{nTry} = Val(40);
+    peakedness = 1.0) where {nTry}
+    lower, upper, mode = promote(lower, upper, mode / 1)
+    T = typeof(lower)
+    mode1 = (mode-lower)/(upper-lower)
+    dln = fit_mode_flat(LogitNormal{T}, mode1, ntry; peakedness)
+    #dln * (upper-lower) + lower
+    Distributions.AffineDistribution{T}(lower, (upper - lower), dln)
+end
+

test/univariate/continuous/logitnormal.jl

@@ -134,3 +134,24 @@ end;
     #plot(d)
     #dln = LogitNormal(0f0,sqrt(2f0)); plot!(dln)
 end;
+
+@testset "shifloNormal_mode Float32" begin
+    d32 = shifloNormal_mode(1.0f0, 3.0f0, 1.5f0)
+    @test partype(d32) == Float32
+    @test mean(d32) isa Float32
+    # wait until fix in AffineDistribution is merged to Distributions.jl
+    # @test rand(d32) isa eltype(d32)
+    m = mode(d32)
+    @test m isa Float32
+    @test isapprox(m, 1.5, atol = 0.1)
+    dmin = minimum(d32)
+    @test dmin == 1.0
+    @test maximum(d32) == 3.0
+    @test scale(d32) == 3 - 1
+    @test scale(d32) isa Float32
+    @test location(d32) == 1.0
+    @test location(d32) isa Float32
+    #using StatsPlots
+    #plot(d32)
+    #dln = LogitNormal(0f0,sqrt(2f0)); plot!(dln)
+end;