Two analytic decisions surfaced by the PR #2 review, both currently encoded one way in config but worth explicit sign-off:
-
PUMF initiation-age floor: 13 vs 8. config.yml sets min: 13 for age_first_cigarette/age_start_daily on PUMF (8 on Master), but SMKG01C_cont/SMKG040_cont carry a 5-11 category (midpoint 8) in all PUMF cycles — the floor of 13 excludes those genuine early initiators from the APC numerator. Note the age spline knots start at 10, which suggests ages below 13 were intended to contribute. Decide: lower the PUMF floor to 8 (consistent with Master), or keep 13 and document why (midpoint-8 category too coarse?).
-
Cycle-1 survey year: 2001 vs 2002. cycle_survey_years had "1": 2002 (and the test agreed) while the inline comment and cycle label said 2001. CCHS 1.1 collected Sept 2000-Nov 2001, so the review changed both to 2001 (commit eeda81d). This shifts cycle-1 cohort assignment (cohort = survey_year - age) by one year. Confirm 2001 is the intended convention, or revert with a documented rationale (e.g. if matching Manuel et al. 2020's SAS convention requires 2002).
Two analytic decisions surfaced by the PR #2 review, both currently encoded one way in config but worth explicit sign-off:
PUMF initiation-age floor: 13 vs 8.
config.ymlsetsmin: 13forage_first_cigarette/age_start_dailyon PUMF (8 on Master), butSMKG01C_cont/SMKG040_contcarry a 5-11 category (midpoint 8) in all PUMF cycles — the floor of 13 excludes those genuine early initiators from the APC numerator. Note the age spline knots start at 10, which suggests ages below 13 were intended to contribute. Decide: lower the PUMF floor to 8 (consistent with Master), or keep 13 and document why (midpoint-8 category too coarse?).Cycle-1 survey year: 2001 vs 2002.
cycle_survey_yearshad"1": 2002(and the test agreed) while the inline comment and cycle label said 2001. CCHS 1.1 collected Sept 2000-Nov 2001, so the review changed both to 2001 (commit eeda81d). This shifts cycle-1 cohort assignment (cohort = survey_year - age) by one year. Confirm 2001 is the intended convention, or revert with a documented rationale (e.g. if matching Manuel et al. 2020's SAS convention requires 2002).