search

lrnv / Copulas.jl

A fully `Distributions.jl`-compliant copula package
https://lrnv.github.io/Copulas.jl/
MIT License
93 stars 10 forks source link

[New feature] Rosenblatt (and inverse) transformations #151

Open lrnv opened 8 months ago

lrnv commented 8 months ago

AnderGray said in Joss's review:

Rossenblat (and inverse) transformation is often used in uncertainty analysis (particularly reliability analysis) to transform random variables into independent standard normal variables.

There is a bit of rosenblatt transformation implemented in BivariateCopulas.jl. I know this transformation is very useful, I am just not really sure how we can make it generic enough to work for all copulas (on one side) and specific enough for standard models to obtain fast computations.

It is also used to do quasi-random-number generation with a dependence structure, which is really useful for some use-cases.

lrnv commented 8 months ago

@AnderGray By the way, would you agree for us to port plotting interface / rosenblatt transforms / other functionalities from BivariateCopulas.jl to Copulas.jl ? We will have to think about how to credit you of course, maybe you should author the PRs

AnderGray commented 8 months ago

@lrnv Just seeing this. Yes, very happy to give it a go :)

lrnv commented 8 months ago

@AnderGray Thanks, I will go in this direction then as soon as bandwidth becomes available again.