Guide overview
The essential idea
A publishable NHANES study begins before code is written. The research question, target population, variable definitions, cycle choice, design specification, and reporting plan should form one auditable chain. This workflow helps prevent a common failure: a polished manuscript whose analytical decisions cannot be reconstructed or defended.
Practical framework
What to apply
Write the protocol map
What will be estimated, compared, or modeled - and why?
- Define the question, estimand, population, exposure, outcome, covariates, subgroups, and time period.
- State primary and secondary analyses before exploring results.
- Justify the selected cycles and whether pooling, stratified estimates, or trend analysis is intended.
- Choose the reporting guideline that matches the study design and planned report.
Build reproducibly
Can another analyst reconstruct the analytic sample and variables?
- Download source files programmatically where practical and preserve cycle, file, and retrieval metadata.
- Maintain a variable dictionary with codebook definitions, units, labels, special codes, and derivations.
- Use scripted joins and recodes; record exclusions and row counts after every material transformation.
- Separate raw, intermediate, analytical, and output files so source data remain unchanged.
Use survey-aware analysis
Do the estimates and standard errors match the design and analysis population?
- Select the correct full-sample or subsample weight using the least-common-denominator principle.
- Specify masked strata and primary sampling units in survey-capable procedures.
- Handle subpopulation analyses with appropriate domain methods rather than deleting everyone outside the subgroup.
- Assess estimate reliability, missingness, influential observations, model assumptions, and sensitivity analyses.
Make the manuscript traceable
Can readers connect each claim to a documented analytical decision?
- Report cycles, component files, eligibility, exclusions, final sample, weights, strata, primary sampling units, and software.
- Describe missing-data decisions, derived measures, combined-cycle weights, and sensitivity analyses.
- Present weighted estimates with uncertainty and show unweighted analytical sample sizes.
- Use proportionate language, cite NCHS documentation, and provide code or a reproducibility statement where possible.
Before you finish
A short quality review
- The research question, estimand, and primary analysis were defined before result review.
- Every reported variable is traceable to a cycle-specific source and transformation.
- The participant flow and final unweighted sample size are reproducible.
- The selected weight and survey design variables are named and justified.
- Reliability, missingness, sensitivity, and subgroup limitations are visible.
- The manuscript follows an appropriate reporting guideline and cites the official NHANES documentation.
Publication begins with design decisions. A good NHANES paper does more than calculate correctly. It shows why the selected cycles, variables, weights, population, and analytical choices answer the stated question.