Add a diagnostic comparing ESS with all iterations and ESS with half of the iterations
Nobody has claimed this yet.
- Dominant language
- R
- Stars
- 171
- Forks
- 26
- Avg merge
- 2d 18h
- Merged PRs (30d)
- 3
Description
Add a diagnostic comparing ESS with all iterations and ESS with half of the iterations (could be average of the first and second half). If CLT holds, then we expect ESS from all iteration to be 2 times the ESS from half of the iterations. For example, Figure 5 in the new Rhat/ESS/MCSE paper https://doi.org/10.1214/20-BA1221 shows an example of problematic cases where Tail-ESS is non-zero but doesn't double when the number of iterations is doubled. The reason for not approximately doubling can be either non-finite variance of the target or convergence issues (as in that Figure 5 as Tail-ESS uses quantiles that have finite variance).
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
No files, tests, or entry points are named. Start by locating the existing ESS diagnostic and its tests, then review how diagnostics handle all versus partial iterations and verify the expected doubling behavior, including the problematic Tail-ESS cases described in the issue.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- r
- Domain
- data
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100