Skip to content

initial setup of insights repository#1

Open
ed-nykaza wants to merge 1 commit into
masterfrom
etn/get-insights-repo-set-up
Open

initial setup of insights repository#1
ed-nykaza wants to merge 1 commit into
masterfrom
etn/get-insights-repo-set-up

Conversation

@ed-nykaza

Copy link
Copy Markdown

No description provided.

@scaubrey scaubrey left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks like a good start and as we add to it, we'll get a better idea of how to structure things.

markjudeconnolly added a commit that referenced this pull request Jun 7, 2026
…8.3 arm sweep + report confirms

Figure layout (developer_note §0 ask #1):
- New utils.plotting.render_4x2_grid helper. Removed the baked-in "Figure X.Xy" in-image prefix from every figure (the report caption is the single source of each figure's number) and tightened the suptitle→panel gap.
- Merged the legible line/bar/Δ-histogram grids and the 5-arm §8.1 per-user violins (8.1b/12.1b) from a grid1/grid2 pair into a single 4×2 PNG (figure_<id>_4x2.png); the denser violins (8.2a, 8.3a/8.3f, 12.3a/12.3c/12.3g) stay split. Updated the four run_test_analysis_8_* figure-count asserts to the merged filenames.

Headline NMA arm → CE=0/BE≤1 (decisions.md D20):
- Switched the featured single NMA arm from CE=0/BE≤∞ to CE=0/BE≤1 wherever one arm is featured — §8.1c and §12.1c (headline_flag) and §8.2 HEADLINE_CLS (the 8.2c CE>0-comparator-line source) — matching §8.4's existing HEADLINE_ARM; the figure titles now name the arm.

§8.3 single-CE=0-arm sweep:
- New figure_8_3a_5way_violin: 8.3a and its Appendix median/rolling twins 12.3a/12.3c now show all 5 day types as two 4×1 grids; 8.3b, 8.3d, 8.3f and 12.3g now feature the explicit CE=0/BE≤1 arm (labelled). Rule: binary violin grids → all 5; tercile grids + histograms + the R-distribution → explicit BE≤1. All §8.3 tables already break out the 3 nested arms (unchanged); §8-supp left on the broadest arm (out of report).
- Housekeeping: removed the dead strata_inf; renamed the builder param strata_inf → ce0_strata; added ce0_label/ce0_flag.

DAY_TYPE_COLORS palette (decisions.md D13 update):
- The 3 nested CE=0 arms now read a Tidepool-brand blue ramp (dark / brand = the BE≤1 headline / light) instead of the green ramp, whose mid step collided with the TIR in-range green; CE>0 grey and HMA bronze unchanged. Applies to every day-type figure; the range-coloured figures keep their endpoint colours. Regenerated for all three cohorts.

D5 reconciliation (decisions.md D5 update):
- Reconciled the figure docstrings (8.1c/12.1c) and the project docs to the D5 update: the CE=0/BE≤1 headline's TIR/TAR/mean-glucose/CV contrasts are method-robust (Method A and Method B concur in sign on all three arms); the divergence caveat attaches to the below-range / hypoglycemia endpoints only. Removed the stale "stringent → descriptive-only / no robust directional claim" framing on the headline; the dated historical records are preserved as superseded and the genuine §8.4 strategy/carb-rate divergence is intact.

Report-ask confirms (developer_note §0):
- Confirmed the §8.3/§12.3 figure-8.3e day-count drop is the intended same-cohort gate (data_loader.restrict_comparator), not a stray filter. Added data_staging/confirm_nma_user_day_age_parity.py to diff a Databricks nma_user_day_age regeneration against the local snapshot. §8.3 figure-letter order kept as-is (filenames trace 1:1 = provenance).

Docs: decisions.md, report_editor_note.md, developer_note.md, architecture.md, project_history.md, todo.md.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants