Claude/fix ard vignette error s h6iq

.Rbuildignore

@@ -13,3 +13,7 @@ manifest.json
 ^\.github$
 ^cran-comments\.md$
 ^CRAN-SUBMISSION$
+^\.claude$
+^CLAUDE\.MD$
+^ENHANCEMENTS\.md$
+^IMPLEMENTATION_PLAN\.md$

.claude/settings.local.json

@@ -0,0 +1,10 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(gh repo view:*)",
+      "Bash(Rscript:*)",
+      "Bash(grep:*)",
+      "Bash(test:*)"
+    ]
+  }
+}

CLAUDE.MD

@@ -0,0 +1,264 @@
+# beeca Package - Claude Agent Context
+
+## Package Overview
+
+**beeca** (Binary Endpoint Estimation with Covariate Adjustment) is a lightweight R package for estimating marginal treatment effects in clinical trials with binary outcomes using logistic regression working models with covariate adjustment.
+
+**Version:** 0.2.0
+**License:** LGPL (>= 3)
+**Authors:** Alex Przybylski, Mark Baillie, Craig Wang, Dominic Magirr (Novartis)
+
+## Purpose and Scope
+
+### Primary Goal
+Facilitate quick industry adoption and use within GxP environments for covariate-adjusted analyses in randomized clinical trials.
+
+### Use Cases
+- Clinical trials with binary endpoints
+- Covariate-adaptive randomization (stratified permuted block or biased coin)
+- Superiority and non-inferiority trials
+- FDA guidance-compliant robust variance estimation
+
+### Summary Measures Supported
+- Risk Difference (diff)
+- Odds Ratio (or)
+- Risk Ratio (rr)
+- Log Odds Ratio (logor)
+- Log Risk Ratio (logrr)
+
+## Methodology
+
+The package implements two robust variance estimation approaches:
+
+1. **Ge et al. (2011)** - Delta method for conditional average treatment effect
+   - Uses sandwich variance estimators (HC0, HC1, HC2, HC3, HC4, HC4m, HC5, model-based)
+   - Suitable for conditional ATE estimation
+
+2. **Ye et al. (2023)** - Robust variance for population average treatment effect
+   - Supports stratification variables
+   - Cross-validated against RobinCar package
+   - Modified implementation option for improved stability
+
+Both methods use g-computation for marginal effect estimation.
+
+## Core Architecture
+
+### Functional Pipeline
+
+The package uses a functional composition pattern with 5 main exported functions:
+
+```r
+glm_object |>
+  predict_counterfactuals(trt) |>    # Generate potential outcomes
+  average_predictions() |>           # Average across population
+  estimate_varcov(method, type) |>   # Robust variance estimation
+  apply_contrast(contrast, ref) |>   # Calculate treatment effect
+  get_marginal_effect()              # Wrapper function
+```
+
+### Key Functions
+
+| Function | Location | Purpose |
+|----------|----------|---------|
+| `get_marginal_effect()` | `R/get_marginal_effect.R` | High-level wrapper orchestrating the pipeline |
+| `predict_counterfactuals()` | `R/predict_counterfactuals.R` | Generate predictions for all treatment levels |
+| `average_predictions()` | `R/average_predictions.R` | Compute mean predictions per treatment |
+| `estimate_varcov()` | `R/estimate_varcov.R` | Estimate variance-covariance matrix |
+| `apply_contrast()` | `R/apply_contrast.R` | Calculate marginal effects with contrasts |
+| `sanitize_model()` | `R/sanitize.R` | Validate glm object (S3 generic) |
+
+### Object Structure
+
+The package augments the standard glm object rather than creating new classes. After processing, a glm object contains:
+
+**Original glm components** (preserved):
+- coefficients, residuals, fitted.values, etc.
+
+**Added by beeca**:
+- `counterfactual.predictions` - Tibble with predictions for all treatment scenarios
+- `counterfactual.means` - Named vector of mean predictions per treatment
+- `robust_varcov` - Variance-covariance matrix with method attribute
+- `marginal_est` - Point estimate(s) of treatment effect(s)
+- `marginal_se` - Standard error(s) of treatment effect(s)
+- `marginal_results` - **Analysis Results Data (ARD) tibble**
+
+## Analysis Results Data (ARD) Structure
+
+The `marginal_results` component is a key output providing pharmaceutical industry-standard reporting format.
+
+### ARD Schema
+
+| Column | Type | Description |
+|--------|------|-------------|
+| `TRTVAR` | character | Treatment variable name |
+| `TRTVAL` | character | Treatment level value |
+| `PARAM` | character | Parameter name (outcome variable) |
+| `ANALTYP1` | character | "DESCRIPTIVE" or "INFERENTIAL" |
+| `STAT` | character | Statistic type (N, n, %, risk, risk_se, contrast, contrast_se) |
+| `STATVAL` | numeric | Numeric value of the statistic |
+| `ANALMETH` | character | Method (count, percentage, g-computation, variance type) |
+| `ANALDESC` | character | Description including beeca version |
+
+### ARD Row Structure
+
+**For 2-arm trial (12 rows):**
+- Rows 1-5: Treatment arm 0 (N, n, %, risk, risk_se)
+- Rows 6-10: Treatment arm 1 (N, n, %, risk, risk_se)
+- Rows 11-12: Contrast estimate and SE
+
+**For 3-arm trial (19 rows):**
+- Rows 1-15: Three arms × 5 statistics each
+- Rows 16-19: Multiple pairwise contrasts
+
+## File Organization
+
+```
+beeca/
+├── R/
+│   ├── get_marginal_effect.R      # Main wrapper function
+│   ├── predict_counterfactuals.R  # Counterfactual generation
+│   ├── average_predictions.R      # Averaging step
+│   ├── estimate_varcov.R          # Variance estimation
+│   ├── apply_contrast.R           # Contrast calculation
+│   ├── sanitize.R                 # Model validation (S3 methods)
+│   ├── trial01.R                  # Example dataset
+│   ├── trial02_cdisc.R           # CDISC example dataset
+│   ├── margins_trial01.R         # SAS comparison data
+│   └── ge_macro_trial01.R        # SAS macro comparison data
+├── tests/testthat/
+│   ├── test-get_marginal_effect.R
+│   ├── test-predict_counterfactuals.R
+│   ├── test-average_predictions.R
+│   ├── test-estimate_varcov.R
+│   ├── test-apply_contrasts.R
+│   └── test-sanitize.R
+├── man/                           # Generated documentation
+├── vignettes/                     # User guides
+├── DESCRIPTION
+├── NAMESPACE
+└── README.md
+```
+
+## Model Requirements
+
+The package validates that input models meet these requirements:
+
+1. **Model Type:** Binomial family with logit link
+2. **Treatment Variable:** Must be a factor with 2+ levels
+3. **Response Variable:** Must be coded as 0/1
+4. **Convergence:** Model must have converged
+5. **Model Matrix:** Must have full rank
+6. **Interactions:** No treatment-covariate interactions allowed
+7. **Missing Data:** No missing values in model data
+
+These are checked via `sanitize_model()` before processing.
+
+## Testing Strategy
+
+- **Test Framework:** testthat (edition 3)
+- **Total Tests:** 80+ test cases
+- **Cross-validation:** Against SAS %margins macro, {margins}, {marginaleffects}, {RobinCar}
+
+Test files mirror the function structure:
+- `test-sanitize.R` - 14 tests for input validation
+- `test-predict_counterfactuals.R` - Counterfactual generation
+- `test-average_predictions.R` - 9 tests including edge cases
+- `test-estimate_varcov.R` - 18 tests (Ge, Ye, stratified)
+- `test-apply_contrasts.R` - 28+ tests for all contrast types
+- `test-get_marginal_effect.R` - End-to-end integration tests
+
+## Dependencies
+
+### Required (Imports)
+- `dplyr` - Data manipulation
+- `sandwich` - Robust variance estimation
+- `stats` - GLM fitting
+- `lifecycle` - Package lifecycle badges
+
+### Suggested (Testing/Vignettes)
+- `testthat` (>= 3.0.0)
+- `knitr`, `rmarkdown`
+- `tidyr`
+- `marginaleffects`, `margins` - Cross-validation
+- `RobinCar` (>= 0.3.0) - Cross-validation
+
+## Example Usage
+
+```r
+library(beeca)
+
+# Prepare data
+trial01$trtp <- factor(trial01$trtp)
+
+# Fit working model and estimate marginal effect
+fit1 <- glm(aval ~ trtp + bl_cov, family = "binomial", data = trial01) |>
+  get_marginal_effect(
+    trt = "trtp",
+    method = "Ye",
+    contrast = "diff",
+    reference = "0"
+  )
+
+# View ARD results
+fit1$marginal_results
+
+# Access specific components
+fit1$marginal_est       # Treatment effect estimate
+fit1$marginal_se        # Standard error
+fit1$counterfactual.means  # Mean predictions per arm
+fit1$robust_varcov      # Variance-covariance matrix
+```
+
+## Documentation Standards
+
+All functions use roxygen2 documentation with:
+- `@description` - Purpose and use case
+- `@details` - Technical details and methods
+- `@param` - Parameter descriptions with constraints
+- `@return` - Return value structure (often using `\tabular`)
+- `@examples` - Executable examples
+- `@references` - Academic citations with DOIs
+- `@seealso` - Links to related functions
+- `@export` or `@keywords internal`
+
+## Quality Assurance
+
+- FDA guidance compliant (2023)
+- Cross-validated against multiple implementations
+- Stable lifecycle badge
+- Continuous integration (R-CMD-check, test-coverage)
+- CRAN release + GitHub development
+
+## References
+
+### FDA Guidance
+FDA. 2023. "Adjusting for Covariates in Randomized Clinical Trials for Drugs and Biological Products."
+https://www.fda.gov/regulatory-information/search-fda-guidance-documents/adjusting-covariates-randomized-clinical-trials-drugs-and-biological-products
+
+### Methodology Papers
+- Ge et al. (2011). "Covariate-Adjusted Difference in Proportions from Clinical Trials Using Logistic Regression and Weighted Risk Differences." Drug Information Journal 45: 481-93. https://doi.org/10.1177/009286151104500409
+
+- Ye et al. (2023). "Robust Variance Estimation for Covariate-Adjusted Unconditional Treatment Effect in Randomized Clinical Trials with Binary Outcomes." Statistical Theory and Related Fields 7(2): 159-63. https://doi.org/10.1080/24754269.2023.2205802
+
+- Magirr et al. (2025). "Estimating the Variance of Covariate-Adjusted Estimators of Average Treatment Effects in Clinical Trials With Binary Endpoints." Pharmaceutical Statistics 24(4): e70021. https://doi.org/10.1002/pst.70021 [PMID: 40557557]
+
+### Related Packages
+- RobinCar: https://cran.r-project.org/package=RobinCar
+- marginaleffects: https://cran.r-project.org/package=marginaleffects
+- margins: https://cran.r-project.org/package=margins
+- sandwich: https://cran.r-project.org/package=sandwich
+
+## Development Resources
+
+- **Package Website:** https://openpharma.github.io/beeca/
+- **GitHub Repository:** https://github.com/openpharma/beeca
+- **Bug Reports:** https://github.com/openpharma/beeca/issues
+- **Vignettes:** `vignette("estimand_and_implementations")`
+
+## Working Group
+
+Developed in collaboration with the ASA-BIOP Covariate Adjustment Scientific Working Group (https://carswg.github.io/), specifically the Software Subteam.
+
+---
+
+**Note for Claude Agent:** This package prioritizes simplicity, GxP compliance, and industry adoption. When proposing enhancements, maintain the lightweight nature, comprehensive testing, cross-validation approach, and pharmaceutical industry standards (ARD format, CDISC compatibility).

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: beeca
 Title: Binary Endpoint Estimation with Covariate Adjustment
-Version: 0.2.0
+Version: 0.3.0
 Authors@R: 
   c(person(given = "Alex", 
            family = "Przybylski", 
@@ -24,31 +24,35 @@ Authors@R:
            email = "dominic.magirr@novartis.com", 
            role = c("aut")
            ))
-Description: Performs estimation of marginal treatment effects for binary 
-  outcomes when using logistic regression working models with covariate 
-  adjustment (see discussions in Magirr et al (2024) <https://osf.io/9mp58/>). 
-  Implements the variance estimators of Ge et al (2011) <doi:10.1177/009286151104500409> 
+Description: Performs estimation of marginal treatment effects for binary
+  outcomes when using logistic regression working models with covariate
+  adjustment (see Magirr et al (2025) <doi:10.1002/pst.70021>).
+  Implements the variance estimators of Ge et al (2011) <doi:10.1177/009286151104500409>
   and Ye et al (2023) <doi:10.1080/24754269.2023.2205802>. 
 Maintainer: Alex Przybylski <alexander.przybylski@novartis.com>
 License: LGPL (>= 3)
 Encoding: UTF-8
 Roxygen: list(markdown = TRUE)
 RoxygenNote: 7.3.2
-Suggests: 
+Suggests:
+    cards,
+    ggplot2,
     knitr,
-    rmarkdown,
-    testthat (>= 3.0.0),
-    tidyr,
     marginaleffects,
     margins,
-    RobinCar (>= 0.3.0)
+    RobinCar (>= 0.3.0),
+    rmarkdown,
+    testthat (>= 3.0.0),
+    tidyr
 Config/testthat/edition: 3
 Depends: 
     R (>= 2.10)
 LazyData: true
-Imports: 
+Imports:
     dplyr,
+    generics,
     lifecycle,
+    rlang,
     sandwich,
     stats
 VignetteBuilder: knitr