fix: warn instead of silently ignoring missingness in recommend_config (#110) - #114
Merged
Merged
Conversation
The Option A trade study's example recommendations span only 23 design points across 3 p-buckets x 4 missingness categories (some cells have a single point) -- too sparse to bucket on responsibly without shipping unreplicated, effectively arbitrary per-cell values. recommend_config() now warns when missingness is passed as anything but the default "auto" instead of accepting and discarding it silently. Closes #110
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
recommend_config()accepted amissingnesskwarg that was silently discarded (# noqa: ARG001 - reserved for future tuning). Per #110, resolved via the issue's option (2): document why it isn't branched on yet and make the no-op explicit, rather than fabricate a missingness-conditional bucketing.Why not branch on it (option 1)?
The Option A trade study's exported example recommendations (
analysis/results/optionA/recommendations.json) span only 23 design points across 3 p-buckets x 4 missingness categories — several (p-bucket, missingness) cells have exactly one point. Bucketing on that would ship single-draw, unreplicated numbers per cell, which is precisely the problem #111 already flags for the existing p-only bucketing (no replication anywhere in the pipeline; sensitivity data doesn't clearly support the shippedhp_vaframing). Doing it properly requires trade-study support for replicated trials (jcm-sci/trade-study#112, still open) before any new baked-in per-cell values could be trusted.Change
recommend_config(missingness=...): passing anything other than the default"auto"now raises aUserWarningpointing at recommend_config accepts a missingness parameter but never uses it #110, instead of accepting and ignoring the value."auto"stays silent (asserted viawarnings.simplefilter("error")); non-default value warns with the expected message.Unreleasedentry.Closes #110
Test plan
just cigreen locally: 481 passed, 90.21% coverage (gate 89%)