We have, Assuming ψ has support on the positive real line,
ξψ=X
Where
X∼Fn and
Fn is the empirical distribution of the data.
Taking the log of this equation we get,
Log(ξ)+Log(ψ)=Log(X)
ξψ
ΨLog(ξ)(t)ΨLog(ψ)(t)=ΨLog(X)
Now, ξ∼Unif[0,1],therefore−Log(ξ)∼Exp(1)
Thus,
ΨLog(ξ)(−t)=(1+it)−1
Given that Ψln(X)=1n∑1000k=1exp(itXk),
With X1...X1000 The random sample of ln(X).
We can now specify completly the distribution of Log(ψ) through its characteristic function:
(1+it)−1ΨLog(ψ)(t)=1n∑k=11000exp(itXk)
If we assume that the moment generating functions of ln(ψ) exist and that t<1 we can write the above equation in term of moment generating functions:
MLog(ψ)(t)=1n∑k=11000exp(−tXk)(1−t)
It is enough then to invert the Moment generating function to get the distribution of ln(ϕ) and thus that of ϕ