[WIP] Implement Proposals.HSlice

[SaranjeetKaur]

No description provided.

[codecov]

Codecov Report

Merging #45 into master will decrease coverage by 0.19%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #45      +/-   ##
==========================================
- Coverage   88.23%   88.03%   -0.20%     
==========================================
  Files           9        9              
  Lines         442      443       +1     
==========================================
  Hits          390      390              
- Misses         52       53       +1     

Flag	Coverage Δ
#unittests	88.03% <ø> (-0.20%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/proposals/Proposals.jl	93.75% <ø> (ø)
src/bounds/Bounds.jl	80.00% <0.00%> (-3.34%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 19f5273...5478297. Read the comment docs.

[mileslucas]

Okay, you need to revert all the whitespace changes you've made to the docstrings and such. Please don't do this again in the future.

[cpfiffer]

I can't really speak to the actual numerical outcome because I'm not super familiar with the methods.

It does seem that HSlice is really long and should be cut up a lot. It also looks like there's lots of repeated code, though I think you've already mentioned you're going to factor this stuff out.

Also, this is a general question for this repo, how is the actual accuracy of these proposals being tested? Like if I give it a simple model, does it get the mean/variance right?

[cpfiffer]

Also, all the tests are failing. Definitely a good idea to fix those before we can make any meaningful comments.

[mileslucas]

Also, this is a general question for this repo, how is the actual accuracy of these proposals being tested? Like if I give it a simple model, does it get the mean/variance right?

Right now I'm testing some trivial properties of the proposals, such as "does the proposed point have higher likelihood than the critical value?" Implicitly it also makes sure we don't hit any of those errors related to extremely inefficient sampling (this is all in test/proposals.jl). In addition, they're being used inside the sampling integration tests, which do have analytical outputs we can test against (this is in test/sampling.jl).

Looking at other nested sampling libraries this actually seems like more coverage already. For example, dynesty's tests are almost all integration tests that don't even test all combos of proposals and bounds.

I'd love to get some more granularity to be able to test the corner cases and expectations of the proposals, but this work isn't directly part of my current research so I can't commit as much time to this package to polish off the corners.