Two papers live here, under different licences. Read this before re-using either.
The paper the whole course replicates, reproduced in ../reproduction/
and worked through in ../examples/.
Kueh MTW, Chew NWS, et al. Body weight categories and fat distribution in relation to all-cause mortality among adults with metabolic dysfunction-associated steatotic liver disease. BMJ Open 2026;16:e113719. doi:10.1136/bmjopen-2025-113719
© Author(s) (or their employer(s)) 2026. Re-use permitted under CC BY-NC. No commercial re-use. Published by BMJ Group.
e113719.full.pdf is redistributed under that licence. CC BY-NC is not CC BY — the
repository’s own CC BY licence does not extend to this file. Full terms:
https://creativecommons.org/licenses/by-nc/4.0/
Its peer review history is not included here, because it carries no stated licence — and it is worth reading, since seeing how professional reviewers critiqued this paper is the closest thing to a worked example of the M2 genre. To get it:
| Published | Our reproduction | |
|---|---|---|
| Analytic N | ~6,300 | 6,371 |
| Groups I / II / III / IV | 386 / 4,541 / 769 / 604 | 391 / 4,585 / 787 / 608 |
| Deaths | 585 | 586 |
| Group IV HR, unadjusted → adjusted | 15.13 → 2.89 | 15.14 → 2.91 |
The gap is documented rather than tuned away — see
examples/M1_replication/.
A fully reproducible NHANES survival analysis with public code, included as a worked example of design-aware practice.
Karim ME, Hossain MB, Zheng C. Examining the Role of Race/Ethnicity and Sex in Modifying the Association Between Early Smoking Initiation and Mortality: A 20-Year NHANES Analysis. AJPM Focus 2025;4(2):100282. doi:10.1016/j.focus.2024.100282
© 2024 The Author(s). Published by Elsevier Inc. Open access under CC BY. https://creativecommons.org/licenses/by/4.0/
Files: PIIS2773065424001007.pdf (article), mmc1.pdf (supplementary content).
Code, analytic data and outputs are archived at Zenodo: https://doi.org/10.5281/zenodo.20764350
Read its Methods alongside
../practice/P4_design_aware_estimate.md. It
states plainly what P4 asks you to do, and what the M2 model critique presses another
group about:
“The design was created on the entire data using the design features: interview weights, clusters, and strata. Subsequently, the authors subset the design to focus on eligible patients to estimate variances using the Taylor series linearization method.”
Subsetting the design object rather than filtering the data before svydesign() is the
difference between correct and too-narrow confidence intervals. Most papers never say
which they did. This one does, and the Zenodo archive lets you check it.
It is also a useful contrast on missing data: the authors decline to impute, and say why — a reliable imputation model could not be built from the available covariates. P5 asks for one justified missing-data analysis. A reasoned refusal is a justification; “under 10% missing, so complete-case is fine” is not.
NHANES microdata and NCHS linked mortality files are US government public-domain data,
not covered by either licence above, and not redistributed here.
reproduction/R/01_download.R fetches them from CDC.