estimator Lindner errors with Nothing

[Datseris]

MWE

x = rand(1000)
y = rand(1000)
z = rand(1000)
ηTf = 1
embedding = EmbeddingTE(; dS = 3, dT = 3, dC = 3, ηTf)
estimator = Lindner(k=10)
independence_test = LocalPermutationTest(TEShannon(; embedding), estimator, nshuffles = 100)
independence(independence_test, x, y, z)

throws

ERROR: MethodError: Cannot `convert` an object of type Nothing to an object of type Float64

Closest candidates are:
  convert(::Type{T}, ::ColorTypes.Gray24) where T<:Real
   @ ColorTypes C:\Users\gd419\.julia\packages\ColorTypes\1dGw6\src\conversions.jl:114
  convert(::Type{T}, ::ColorTypes.Gray) where T<:Real
   @ ColorTypes C:\Users\gd419\.julia\packages\ColorTypes\1dGw6\src\conversions.jl:113
  convert(::Type{T}, ::Union{Static.StaticBool{N}, Static.StaticFloat64{N}, Static.StaticInt{N}} where N) where T<:Number
   @ Static C:\Users\gd419\.julia\packages\Static\dLrtk\src\Static.jl:414
  ...

Stacktrace:
 [1] setindex!(A::Vector{Float64}, x::Nothing, i1::Int64)
   @ Base .\array.jl:969
 [2] permuted_Îs(X::StateSpaceSet{1, Float64}, Y::StateSpaceSet{3, Float64}, Z::StateSpaceSet{6, Float64}, measure::CMIShannon{Shannon{Int64}}, est::Lindner{Int64}, test::LocalPermutationTest{TEShannon{Shannon{Int64}, EmbeddingTE}, Lindner{Int64}, CausalityTools.NeighborCloseness, TaskLocalRNG})
   @ CausalityTools C:\Users\gd419\.julia\dev\CausalityTools\src\independence_tests\local_permutation\LocalPermutationTest.jl:201     
 [3] independence(::LocalPermutationTest{TEShannon{Shannon{Int64}, EmbeddingTE}, Lindner{Int64}, CausalityTools.NeighborCloseness, TaskLocalRNG}, ::Vector{Float64}, ::Vector{Float64}, ::Vector{Float64})
   @ CausalityTools C:\Users\gd419\.julia\dev\CausalityTools\src\independence_tests\local_permutation\transferentropy.jl:23
 [4] top-level scope
   @ c:\Users\gd419\OneDrive - University of Exeter\Projects\CloudControlingRTSA\_research\ccf_basic_te_analysis.jl:68

Using estimator = FPVP() instead works.

[Datseris]

I also get error for Zhu1. I don't understand. Is this because of my pr #343 ? Tests pass there and I didn't really change the estimation code.

[Datseris]

Additionally, I would say the documentation around Lindner can be improved. It doesn't refer to a conditional timeseries and its documentation says that it computes $TE(X→Y)$. But then if I use it with LocalPermutationTest I must have a conditional timeseries.

[kahaaga]

Good catch! This is not an error on your part. It is an actual bug. It happens because for all the low-level estimators (entropy-based, mutual-info based, CMI-based), the cmi = te_to_cmi(measure) line in independence_tests/local_permutation/transferentropy.jl converts the incoming TEShannon measure to a CMIShannon measure. This is a trick that avoids some code duplication internally, but it fails to consider that the estimator might be defined only for TEShannon, not for CMIShannon.

This is a quick fix.