MLE for GEV distribution

[juliohm]

It would be great to have maximum likelihood fits implemented for the generalized extreme value family. I can try tackle it myself if I find the time, but if someone has done it in the past, please let me know and I can try reuse the code.

[juliohm]

Would it be ok to make Optim.jl a dependency for this or is it too much?

[andreasnoack]

I don't think we should have general support for fitting in this package so I'd say no. It is better to have estimation specific functionality elsewhere.

[juliohm]

Agree, where should estimation functionality exist?

Next quarter I will be a teaching assistant and I was hoping to show students a simple fit(GEV, data). If there is a package where these things are being developed, I can submit a PR there instead.

The issue with GEV specifically is that it requires constrained optimization, and Optim.jl would come in handy as a dependency. I don't want to reinvent the wheel.

[andreasnoack]

I agree that Optim.jl would be a natural dependency for such a package. I don't know if there are any current efforts to generally support fitting distributions beyond what is in this package. Maybe a new package should be initiated.

[juliohm]

@andreasnoack can this package be initiated in the organization by some of you? I can submit PRs there instead.

[juliohm]

Thinking more carefully about it, I think extreme value theory deserves a separate package. I will provide the students with Jupyter notebooks this quarter, and will try to organize the material in a package in the future.