Improvements to ess_rhat

[sethaxen]

This PR implements the nonbreaking parts of #22 (specifically, the proposal in #22 (comment)):

• makes splitting of chains optional
• support passing an estimator to compute an estimate-specific ESS (and R-hat too, though I don't think this is commonly used)
• Add ess_rhat_bulk to compute rank-normalized ESS and R-hat
• Add ess_tail and rhat_tail (the two cannot be simultaneously computed since they use different transformations)

Currently this is marked as a draft while I write more tests and try to figure out how improve type-inferrability for utility functions.

[coveralls]

Pull Request Test Coverage Report for Build 3883137860

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 60 of 63 (95.24%) changed or added relevant lines in 2 files are covered.
• 5 unchanged lines in 2 files lost coverage.
• Overall coverage decreased (-0.5%) to 94.421%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/ess.jl	33	36	91.67%

Files with Coverage Reduction	New Missed Lines	%
src/utils.jl	2	97.33%
src/ess.jl	3	95.97%

Totals
Change from base Build 3698624468:	-0.5%
Covered Lines:	660
Relevant Lines:	699

---

💛 - Coveralls

[codecov]

Codecov Report

Base: 94.91% // Head: 95.19% // Increases project coverage by +0.27% 🎉

Coverage data is based on head (a090c4b) compared to base (57cacc1).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #53      +/-   ##
==========================================
+ Coverage   94.91%   95.19%   +0.27%     
==========================================
  Files          10       10              
  Lines         669      708      +39     
==========================================
+ Hits          635      674      +39     
  Misses         34       34              

Impacted Files	Coverage Δ
src/ess.jl	97.98% <100.00%> (+0.11%)	⬆️
src/utils.jl	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[devmotion]

I haven't checked the PR in detail yet (since I'm on my phone and tests seem to fail) but one immediate suggestion is to avoid the JLD dependency and the JLD file. Either just use a text file or even better create the data in the tests without storing it.

[sethaxen]

...tests seem to fail

yes, the failures are in rstar, which is not touched here, so I'm guessing it's an issue with Random seed.

...one immediate suggestion is to avoid the JLD dependency and the JLD file. Either just use a text file or even better create the data in the tests without storing it.

Generating in the tests is cumbersome. When I tried sampling with Turing, I gave up after nearly an hour of sampling. StanSample can do it within a few minutes, but then we need CmdStan installed to run the tests locally and globally. Plus I wouldn't assume the tests would pass for all possible seeds to the sampler, so if CmdStan ever made a change to its sampler, our tests could suddenly begin failing. I'll add these draws as a text file.

[sethaxen]

I've implemented all requested changes. Only the windows tests fail, but that's expected. (see #56 (comment))

The references now look a little strange, but it seems to be the best we can do right now:

[devmotion]

Very nice, great PR! Looks good to me once the Linux test passes.

[sethaxen]

Very nice, great PR! Looks good to me once the Linux test passes.

Thanks for the review! Can you merge? The repo settings require all tests pass before merge, and I don't have permissions to override.

[devmotion]

I changed the settings, now PRs can be merged (from everyone with merge permissions) even if Windows tests fail.

[sethaxen]

Thanks!