Phase 0: WebR Package Validation Framework & Documentation

.gitignore

@@ -1,2 +1,11 @@
 __pycache__/
 *.pyc
+
+# R
+.Rhistory
+.RData
+.Rproj.user/
+*.Rproj
+
+# Shinylive cache (large binary assets, downloaded at export time)
+.shinylive/

Automated_Methods_and_Results_Drafting.md

@@ -0,0 +1,177 @@
+# Optional Step: Automated Drafting of Methods and Results for Peer-Reviewed Publications
+
+## Overview
+
+This optional step allows users to generate a **draft Methods and Results section**, written in the style of a peer-reviewed journal article, based strictly on:
+
+- Information provided by the user about the study context and design
+- The statistical methods actually executed by the app
+- The numerical results, tables, and plots produced by the app
+- The exact R commands used to perform the analyses
+
+The aim is **not** to replace scientific judgment or authorship, but to:
+- Save time on first drafts
+- Improve consistency and completeness
+- Ensure alignment between analyses, code, and written reporting
+- Support reproducible, transparent reporting
+
+The generated text is intended as a **draft for review, editing, and approval by the user**.
+
+---
+
+## User-Provided Study Information
+
+Before generating text, the user is prompted to complete a short structured form capturing information that **cannot be inferred from the data alone**.
+
+### Required Fields
+
+The user will be asked to provide:
+
+#### 1. Study Design
+Examples:
+- Randomised controlled trial (individual or cluster)
+- Cohort study
+- Cross-sectional study
+- Case–control study
+- Interrupted time series
+- Quasi-experimental study
+
+#### 2. Study Setting
+Examples:
+- Country or region
+- Healthcare, educational, community, or other setting
+- Time period of data collection (if applicable)
+
+#### 3. Participants
+- Inclusion criteria
+- Exclusion criteria
+- Any relevant eligibility restrictions
+
+#### 4. Timepoint Definitions
+- Definition of baseline
+- Follow-up timepoints
+- Primary analysis timepoint
+- Any secondary or sensitivity analysis timepoints
+
+#### 5. Outcome Definitions
+For each outcome:
+- Outcome name (exact name to be used in the manuscript)
+- Outcome type (continuous, binary, count, time-to-event)
+- Measurement instrument or source (if relevant)
+- Direction of effect (higher = better/worse, if relevant)
+- Primary vs secondary outcome designation
+
+All user inputs are treated as **authoritative** and are not modified or inferred by the system.
+
+---
+
+## Automatically Captured Analysis Information
+
+The app automatically constructs a structured **analysis manifest** containing:
+
+### Statistical Methods
+- Descriptive statistics produced
+- Statistical models fitted
+- Model formulae
+- Link functions and distributions
+- Covariate adjustment sets
+- Random effects or clustering specifications
+- Robust standard errors or variance estimators
+- Multiple imputation or missing data handling methods (if used)
+- Sensitivity or secondary analyses run
+
+### Software and Code
+- R version
+- Package names and versions
+- Exact R commands executed (or a reproducible template populated with user-specific choices)
+
+### Results
+- Sample sizes and flow counts
+- Summary statistics
+- Model estimates, standard errors, confidence intervals, and p-values
+- Generated tables
+- Generated plots and figure captions
+
+No raw individual-level data are sent to the language model.
+
+---
+
+## LLM Drafting Step
+
+### Inputs to the LLM
+
+The language model receives **only**:
+- The user-provided study information
+- The structured analysis manifest
+- The generated tables and plots
+- The executed R commands
+
+### Outputs from the LLM
+
+The model generates:
+- A **Methods section**, including:
+  - Study design and setting
+  - Participants and eligibility criteria
+  - Outcome definitions
+  - Statistical analysis methods
+  - Software and reproducibility statement
+- A **Results section**, including:
+  - Participant flow and sample characteristics
+  - Primary and secondary outcome results
+  - In-text references to tables and figures
+  - Descriptions of plots consistent with journal conventions
+
+The writing style is neutral, precise, and appropriate for submission to a peer-reviewed journal.
+
+---
+
+## Hallucination Prevention Instruction (Mandatory)
+
+The following instruction is explicitly included in every LLM prompt:
+
+> **Important instruction to the language model**
+>
+> - Use *only* the information provided in the structured inputs.
+> - Do **not** invent, infer, assume, or embellish any methods, results, study characteristics, or interpretations.
+> - Do **not** introduce statistical methods, analyses, covariates, or outcomes that are not explicitly listed.
+> - Do **not** report values, sample sizes, p-values, confidence intervals, or effect estimates that are not present in the supplied results.
+> - If required information is missing, state this explicitly (e.g. “The study setting was not specified”).
+> - Do not add causal language unless explicitly appropriate for the stated study design.
+> - Do not include a discussion or interpretation of findings beyond factual reporting.
+
+The model’s role is strictly **translation from structured truth to narrative text**, not interpretation or inference.
+
+---
+
+## User Review and Export
+
+After generation:
+- The draft Methods and Results sections are displayed to the user
+- All tables and plots are shown alongside the text
+- Users can copy, download, or export the content (e.g. Markdown, Word, or Quarto)
+- The user remains fully responsible for:
+  - Accuracy
+  - Interpretation
+  - Journal-specific formatting
+  - Final approval of the manuscript text
+
+---
+
+## Intended Use and Limitations
+
+- The generated text is a **draft**, not a final manuscript
+- It does not replace domain expertise, peer review, or editorial judgment
+- It is designed to support reproducible reporting and reduce clerical burden
+- It is especially suitable for:
+  - Standard regression analyses
+  - Pre-specified analysis plans
+  - Grant, protocol, or manuscript drafting workflows
+
+---
+
+## Summary
+
+This optional step provides a controlled, transparent way to transform:
+**data → analysis → results → publication-style reporting**
+
+By combining user-supplied study context with machine-captured analytical truth, the system enables fast, accurate, and reproducible drafting of Methods and Results sections while explicitly guarding against hallucination or over-interpretation.