This package provides an implementation of Yeo Johnson transformation. See Wikipedia - Power Transform for more information.
Requires Julia 1.0 or higher version.
This package has not registered yet.
] add https://github.com/tk3369/YeoJohnsonTrans.jl
The simplest way is to just call the transform function with an array of numbers.
julia> using Distributions, UnicodePlots, YeoJohnsonTrans
julia> x = rand(Gamma(2,2), 10000) .+ 1;
julia> histogram(YeoJohnsonTrans.transform(x))
┌────────────────────────────────────────┐
(0.6,0.8] │▇▇ 168 │
(0.8,1.0] │▇▇▇▇▇▇▇▇▇▇▇▇ 837 │
(1.0,1.2] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1586 │
(1.2,1.4] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 2089 │
(1.4,1.6] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 2307 │
(1.6,1.8] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1816 │
(1.8,2.0] │▇▇▇▇▇▇▇▇▇▇▇▇▇ 914 │
(2.0,2.2] │▇▇▇▇ 253 │
(2.2,2.4] │ 28 │
(2.4,2.6] │ 2 │
└────────────────────────────────────────┘
You can examine the power transform parameter (λ) dervied by the program:
julia> YeoJohnsonTrans.lambda(x).value
-0.20252181252892174
You can transfrom the data using your own λ:
julia> histogram(YeoJohnsonTrans.transform(x, -0.3))
┌────────────────────────────────────────┐
(0.6,0.7] │▇ 49 │
(0.7,0.8] │▇▇▇▇▇ 204 │
(0.8,0.9] │▇▇▇▇▇▇▇▇▇▇▇ 427 │
(0.9,1.0] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 703 │
(1.0,1.1] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 916 │
(1.1,1.2] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1103 │
(1.2,1.3] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1287 │
(1.3,1.4] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1376 │
(1.4,1.5] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1354 │
(1.5,1.6] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1125 │
(1.6,1.7] │▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 773 │
(1.7,1.8] │▇▇▇▇▇▇▇▇▇▇▇ 432 │
(1.8,1.9] │▇▇▇▇▇ 197 │
(1.9,2.0] │▇ 45 │
(2.0,2.1] │ 8 │
(2.1,2.2] │ 1 │
└────────────────────────────────────────┘