use posterior::gpdfit and posterior::qgeneralized_pareto()

[avehtari]

Since we do use posterior package elsewhere,
replace gpdfit() and qgpd() with posterior::gpdfit and posterior::qgeneralized_pareto()

This will make maintenance easier. For example, there is a PR for posterior::gpdfit() to make it more robust and posterior::qgeneralized_pareto() already is more robust than qgpd()

This is related to #249, but timing is due to the fix PR for posterior::gpdfit()

[jgabry]

I think the R cmd check error is because the NAMESPACE file is still trying to export gpdfit from loo. If you run devtools::document() it should fix this and then we can see if all the tests pass.

[avehtari]

If it has been exported, then someone might be using it? Should I put it back and define it with posterior::qgeneralized_pareto()?

[avehtari]

At least in github only I and Juho Timonen have used loo::gpdfit, so I guess it can be removed

[jgabry]

Good point. Yeah you can just replace the bodies of the functions with calls to the posterior versions.

[jgabry]

At least in github only I and Juho Timonen have used loo::gpdfit, so I guess it can be removed

OK in that case, we can try removing it and we will find out in the reverse dependency checks whether any packages are using it

[jgabry]

The error now is because apparently these functions aren't exported from posterior, they're just internal functions. That would be easy to change but it would require doing a CRAN release of posterior.

[avehtari]

my bad, I guess this needs to wait for the next posterior release

[jgabry]

There is also #290 from @VisruthSK that is waiting to be merged until there is a new release of posterior

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.71%. Comparing base (23f2117) to head (dce7e3c).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #305      +/-   ##
==========================================
- Coverage   92.78%   92.71%   -0.07%     
==========================================
  Files          31       30       -1     
  Lines        2992     2965      -27     
==========================================
- Hits         2776     2749      -27     
  Misses        216      216              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jgabry]

I think this needs to bump the posterior version in DESCRIPTION (unless it's merged after the other one that does that, but not sure which will be merged first)

[VisruthSK]

Right, in my head it was being merged after the previous one. Will fix that now.

[jgabry]

Right, in my head it was being merged after the previous one. Will fix that now.

We need to figure out what to do about the warnings in the other PR, so will probably do this one first. But for this one, it occurs to me that we need to reexport gpdfit from posterior in order to avoid breaking backwards compatibility (it's currently exported from loo so could be called as loo::gpfit in someone's code.

[VisruthSK]

Will restore tests too then

[jgabry]

Do you think we need to test it in addition to exporting it? I assume it's tested in posterior. Maybe good to test here too, but I'm not sure

[VisruthSK]

If we are rexporting it I think it makes to test, mainly so that the API is the same no? The tests were super minimal anyway.

[jgabry]

Ok yeah good call