BiostatAgent

Comprehensive Biostatistics Agent Ecosystem -- 30 specialized agents, 17 workflow commands, and 45 methodology skills for R-based statistical analysis

A unified Claude Code plugin marketplace consolidating four specialized biostatistics plugins:

Quick Start

1. Add the Marketplace

/plugin marketplace add choxos/BiostatAgent

2. Install Plugins

Install all plugins or select specific ones:

# Install all
/plugin install bayesian-modeling itc-modeling r-tidy-modeling clinical-trial-simulation

# Or install individually
/plugin install bayesian-modeling
/plugin install itc-modeling
/plugin install r-tidy-modeling
/plugin install clinical-trial-simulation

3. Install R Dependencies

# Core dependencies (install as needed)
install.packages(c(
  # Bayesian modeling
  "rstan", "R2jags", "R2WinBUGS", "bayesplot", "loo",
  # ITC/NMA
  "meta", "netmeta", "gemtc", "multinma", "maicplus",
  # Tidy modeling
  "tidyverse", "tidymodels", "recipes", "parsnip", "workflows",
  # Clinical trials
  "simtrial", "Mediana", "gsDesign2", "survival"
))

# CmdStanR is distributed from the Stan R-universe, not CRAN.
install.packages("cmdstanr", repos = c("https://stan-dev.r-universe.dev", getOption("repos")))
cmdstanr::install_cmdstan()

Plugins Overview

1. Bayesian Modeling (bayesian-modeling)

Create, review, and validate Bayesian models across four languages:

• Stan via cmdstanr -- Modern HMC/NUTS sampling with the current installed CmdStan
• PyMC -- Python-native Bayesian modeling with ArviZ diagnostics
• JAGS -- Cross-platform Gibbs sampling with R2jags
• WinBUGS -- Classic BUGS implementation with R2WinBUGS

Commands:

Agents: model-architect, stan-specialist, pymc-specialist, bugs-specialist, model-reviewer, test-runner

Skills: stan-fundamentals, pymc-fundamentals, bugs-fundamentals, hierarchical-models, regression-models, time-series-models, survival-models, meta-analysis, model-diagnostics

2. ITC Modeling (itc-modeling)

Expert agents for indirect treatment comparison following NICE DSU guidance:

• Pairwise Meta-Analysis -- Fixed/random effects with meta, metafor, bayesmeta
• Network Meta-Analysis -- Frequentist (netmeta) and Bayesian (gemtc)
• MAIC -- Matching-adjusted indirect comparison with maicplus
• STC -- Simulated treatment comparison
• ML-NMR -- Multilevel network meta-regression with multinma

Commands:

Agents: itc-architect, pairwise-meta-analyst, nma-specialist, maic-specialist, stc-specialist, ml-nmr-specialist, itc-code-reviewer

Skills: tidy-itc-workflow, pairwise-ma-methodology, nma-methodology, maic-methodology, stc-methodology, ml-nmr-methodology

3. R Tidy Modeling (r-tidy-modeling)

Comprehensive R data science and biostatistics following tidyverse and tidymodels best practices:

• Data Wrangling -- dplyr, tidyr, data transformation
• Feature Engineering -- recipes, preprocessing, transformations
• Model Building -- parsnip, workflows, tidymodels
• Visualization -- ggplot2, publication-ready figures
• Reporting -- Quarto, R Markdown, reproducible reports
• Biostatistics -- Clinical trials, survival analysis, epidemiology, meta-analysis, genomics, pharmacokinetics, health economics, and more

Commands:

Agents: r-data-architect, tidymodels-engineer, feature-engineer, biostatistician, data-wrangler, viz-specialist, reporting-engineer, r-code-reviewer, r-docs-architect, r-tutorial-engineer

Skills (23):

4. Clinical Trial Simulation (clinical-trial-simulation)

Design and simulate clinical trials using simtrial and Mediana:

• simtrial -- Time-to-event simulations, weighted logrank, MaxCombo
• Mediana -- Clinical Scenario Evaluation, multiplicity, Word reports
• gsDesign2 -- Group sequential designs, alpha spending

Commands:

Agents: simulation-architect, tte-specialist, cse-specialist, multiplicity-expert, gs-design-specialist, power-optimizer, code-reviewer

Skills: simtrial-fundamentals, mediana-fundamentals, multiplicity-methods, time-to-event-methods, group-sequential-methods, power-optimization-patterns, clinical-trial-design-patterns

Repository Structure

BiostatAgent/
├── .claude-plugin/
│   └── marketplace.json              # Unified plugin manifest
├── plugins/
│   ├── bayesian-modeling/
│   │   ├── agents/                   # 6 agents
│   │   ├── commands/                 # 3 commands
│   │   └── skills/                   # 9 skills
│   ├── itc-modeling/
│   │   ├── agents/                   # 7 agents
│   │   ├── commands/                 # 2 commands
│   │   └── skills/                   # 6 skills
│   ├── r-tidy-modeling/
│   │   ├── agents/                   # 10 agents
│   │   ├── commands/                 # 7 commands
│   │   └── skills/                   # 23 skills
│   └── clinical-trial-simulation/
│       ├── agents/                   # 7 agents
│       ├── commands/                 # 5 commands
│       └── skills/                   # 7 skills
├── README.md
└── LICENSE

Usage Examples

Bayesian Modeling

Create a hierarchical model for patient outcomes nested within hospitals.
Use Stan with cmdstanr.

ITC Analysis

I have IPD for trial A and AgD for trial B.
Help me run a MAIC to compare treatments.

Tidy R Analysis

Build a predictive model for patient readmission using tidymodels.
Include cross-validation and hyperparameter tuning.

Clinical Trial Simulation

Calculate power for a survival trial with HR=0.7, 300 events, alpha=0.025.
Use weighted logrank for non-proportional hazards.

All Components

Agents (30 total)

Commands (17 total)

Skills (45 total)

Contributing

See CONTRIBUTING.md for how to add agents, commands, or skills. Run python3 scripts/validate-marketplace.py before opening a PR — the same validator runs in CI to guarantee .claude-plugin/marketplace.json stays in sync with the filesystem.

Release notes live in CHANGELOG.md.

License

MIT License -- see LICENSE for details.

Acknowledgments

• Stan Development Team for Stan
• PyMC Developers for PyMC
• Merck & Co. for simtrial
• Mediana Inc. for Mediana
• tidymodels team for tidymodels
• Anthropic for Claude Code