Add relational2 dataset to A/B test the SATA option-as-answer representation#7
Open
jacobmpeters wants to merge 3 commits into
Open
Add relational2 dataset to A/B test the SATA option-as-answer representation#7jacobmpeters wants to merge 3 commits into
jacobmpeters wants to merge 3 commits into
Conversation
Supports BOTH Select-All-That-Apply representations until we decide: - base `responses` fact stays as-is (legacy "option-as-question": each checked option is its own row, option = question_concept_id, parent = source_question_concept_id) - new view `v_responses_sata_v2` presents "option-as-answer" (question = the SATA parent, chosen option = response_value_*), matching how single-select Multiple Choice is already modeled. It's a pure relabeling view over `responses` — no data copied, fully reversible via orig_* lineage columns + a `sata_remodeled` flag. Scope = SATA only: grids are NOT remodeled (a grid sub-item is a genuinely distinct question) and MC single-select is already option-as-answer. SATA is identified by question_type LIKE '%select all that apply%' (isolated in one CTE; depends on dictionary question_type quality — see the doc). Downstream consumers (incl. the OMOP hash) choose a source by reading either `responses` or `v_responses_sata_v2`. NOTE: SATA rows hash differently between the two (3 of the 4 hashed fields move), so pick the representation before freezing the hash contract. Adds docs/sata_representation.md and a production-free smoke test that runs the real view SQL in DuckDB (SATA remodeled; grid/MC/free-text unchanged; lossless).
…e_id on responses)
Replaces the earlier in-place view approach with a second dataset, relational2, so the alternative Select-All-That-Apply representation can be evaluated against real data side-by-side with relational (which stays untouched). - sql/relational2/build_responses.sql: creates the relational2 dataset and relational2.responses — a controlled A/B copy of relational.responses (a transform, not a re-unpivot) where SATA rows are remodeled option-as-answer (question <- parent, answer <- option, source_question <- NULL) and response_unique_id is RECOMPUTED via the UDF on the new fields. Non-SATA rows are byte-identical, so any downstream difference is attributable only to the SATA remodel. Scope = SATA only (grids/MC untouched). - sql/relational2/compare_with_relational.sql: quantifies the difference (rows, ids changed, id health) — no verbatim values. - scripts/smoke_test_relational2.py: synthetic test proving grain preserved, SATA remodeled, non-SATA + their ids unchanged, SATA ids recomputed & differ, all ids in OMOP range. Passes. - docs/sata_representation.md rewritten around relational2 (incl. the dim implication: SATA parents would move into the question dim on enrichment). - Remove the superseded v_responses_sata_v2 view + its smoke test. Brings feature/sata-dual-representation up to main (UDF, response_unique_id on responses) via merge.
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.
Adds a second dataset,
relational2, to evaluate the alternative Select-All-That-Apply representation against real data without disturbingrelational.sql/relational2/build_responses.sqlbuildsrelational2.responsesas a controlled A/B copy ofrelational.responses(a transform, not a re-unpivot): SATA rows are remodeled option-as-answer (question ← parent,answer ← option,source_question ← NULL) andresponse_unique_idis recomputed via the UDF on the new fields. Non-SATA rows are byte-identical, so any difference is attributable only to the SATA remodel. Scope = SATA only (grids/MC untouched).compare_with_relational.sql— quantifies the difference (rows, ids changed, id health); no verbatim values.scripts/smoke_test_relational2.py— synthetic, passes (grain preserved; SATA remodeled; non-SATA + ids unchanged; SATA ids recomputed & differ; all ids in OMOP range).docs/sata_representation.mdrewritten aroundrelational2, including the dim implication (SATA parents would move into thequestiondim on enrichment).v_responses_sata_v2view + its smoke test (superseded).Brings the branch up to
main(UDF,response_unique_idon responses). No production data read.