All issues

Search results

Eleven smaller review observations (M7-M11, M14, L1-L7): reproducibility, validation gaps, docs-vs-code

Status: Open.

#230 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

prod revision_aware grid row uses g_quantreg (quantile_reg default method), but the explore config that validated the family uses quantreg_fn (method = "fn") — explore-vs-prod divergence

Status: Open.

#228 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

covid explore's nhsn_archive keeps the nation as a geo_type = state row ("usa" vs the flu twin's usa → us rename)

Status: Open.

#227 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

nwss_hhs_region aggregation computes region_value = mean(region_value * population) / sum(population) — off by a factor of the region's state count, and NA-prone

Status: Open.

#226 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

data_substitutions drops unmatched correction rows without any warning

Status: Open.

#225 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

parse_prod_weights silently reuses the oldest ("defaults") block when the requested date has no block, defeating the downstream empty-table abort

Status: Open.

#224 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

climatological_model's week-window filter uses plain abs() on epiweeks 1–53 — no wraparound around the year boundary (the sibling climate_median path already uses mod_dist(..., 53))

Status: Open.

#223 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

run_prod_if_fresh.R completes green and creates the done-marker even when g_submission_directory silently defaulted to "cache" (no submission files written)

Status: Open.

#222 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

S3 targets-store sync (make pull) is followed by unverified RDS/qs/parquet deserialization — treat the stores as an untrusted input boundary

Status: Open.

#221 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

population/state-code/HHS crosswalk CSVs are fetched from refs/heads/main at run time on the submission path, and location typing relies on readr's implicit column guessing

Status: Open.

#220 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

score_forecasts rescales wk inc covid prop ed visits by ×100 but not wk inc flu prop ed visits, so flu NSSP scores compare proportion-scale forecasts against percent-scale truth

Status: Open.

#219 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026

scripts/oracle/compare.R: NA/NaN differences are dropped by na.rm = TRUE, so NA-introducing (or NA-repairing) refactors pass the golden gate as "match"

Status: Open.

#218 In cmu-delphi/exploration-tooling;

· docxology opened on Sep 16, 2026