Overview | 1. Download the files | 2. Read column positions from the SAS program | 3. Read only the columns you need | The same steps with the package functions | 4. Missing-value codes | 5. Look up variables in the codebook | 6. Smoking status, checked against cotinine | 7. Link mortality | 8. Survey-weighted Cox model | Notes

Overview | Package options | Background: NHANES structure | 1. What cycles and files are available? | 2. Discover variables | 3. Download laboratory data | 4. Download questionnaire data | 5. Harmonize across cycles | 6. Recode questionnaire variables | 7. Stack demographics and merge all components | 8. Link mortality and prepare the survival dataset | 9. Survey-weighted Cox model | Choosing the correct survey weight | Pooling across cycles | Notes on data management | Cross-cycle variable harmonization | SEQN is a character identifier | NHANES questionnaire coding | Notes on the public-use LMF

Background and Motivation | Why not just use cph()? | Why not just use svycoxph()? | The fusion approach | Key references | Step 1: Examine object structures | Comparing weighted and unweighted estimates | Step 2: Identify Design slot dependencies | Step 3: Implement the fusion | Comparing anova.rms() output: survey-correct vs naive | Step 4: Survey-weighted baseline hazard | Step 5: Degrees of freedom correction | Conclusions and current limitations | What works | Empirical demonstration of design effects | When survey weighting is and is not necessary | Current limitations | The nhanesR package | A broader methodological agenda | Session information

What nhanesR does | Installation | Setup and options | Default cache location | Permanent configuration via .Rprofile | Checking and changing settings interactively | Function map | Stage 1 — Discovery | Stage 2 — Variable search | Stage 3 — Download | Stage 4 — Harmonize and stack | Stage 5 — Mortality linkage | Stage 6 — Survival analysis preparation | Cache management | Typical workflow | Further reading

Motivation | Key references | Notation | Cox: the partial-likelihood contribution | Piecewise exponential: the Poisson-trick contribution | Profiling out $\alpha_k$ recovers Cox | mgcv-Poisson (PAMM): the same contribution, a penalized $\alpha_k$ | Side-by-side comparison | Design weights: why the effect on β differs from the effect on marginal means | The marginal case: weighted mean as an intercept-only estimating equation | The conditional case: Binder's (1992) weighted partial likelihood | Why the numerator doesn't need weighting but the denominator does | A runnable numeric demonstration | Convergence of PEM toward Cox as $K \to \infty$ | Baseline cumulative hazard: three estimates, one true curve | Design weights: conditional robustness vs. marginal sensitivity | Discussion | Current limitations of this document | Session information

The problem | The quality gap in published NHANES research | Two structural barriers to full-cycle, properly-weighted analyses | A pragmatic, hypothesis-first approach | The analytic panel | The CDC catalog gap for 1999–2002 BIOPRO data | Step 1: Discovering where an analyte lives | Example 1: A well-behaved analyte | Example 2: A variable with a prefix change | Example 3: An analyte missing from one cycle | Example 4: An analyte confined to a few cycles | Example 5: A questionnaire variable with a case change | Step 2: Downloading with automatic harmonisation | Downloading ALP across all available cycles | Downloading ALT across all available cycles | Downloading AST across its available cycles | Downloading the questionnaire variable with a case change | Example 6: Disambiguation by component — serum versus urine albumin | Step 3: Assembling the multi-analyte base | Step 4: Derived variables | What the manual approach looks like | Using nhanesR with AI coding assistants | Further reading | Summary of functions used | Session information

Overview | Step 1 — Locate the variables | Step 2 — Download and stack | Step 3 — Calculate UACR | Step 4 — Method comparability and bridging | Known method changes | Bridging equations for serum creatinine | Step 5 — Survival analysis with survey weights | Notes and caveats | See also