Multiplicity and Graphical Procedures (bioSkills)

A Claude Code skill for designing the multiplicity strategy of a confirmatory trial: allocating α across primary, key-secondary and subgroup families, and choosing a procedure that actually controls the familywise error rate under the dependence you have.

   
Type Claude Skill
Supplier GPTomics bioSkills (community OSS, MIT)
Availability GA — part of the bioSkills collection
Pricing Free / OSS (MIT) — gMCP, graphicalMCP, gatekeeping, multcomp, multxpert (R) and statsmodels (Python) are separately installed OSS.
Capabilities Read/Write — Claude runs the skill’s workflow locally (R, with a Python fallback), not as an MCP tool
Verified works · 2026-08-13
Security cleared · 2026-08-13 — GPTomics/bioSkills MIT, no external credentials

How to install

bioSkills is not an npm package — skills are plain markdown/code read directly by the agent. Clone the repo, then either run the installer for the whole category or copy the single skill directory.

  • Claude Code — clone and install via the bundled script: 
    git clone https://github.com/GPTomics/bioSkills
cd bioSkills
./install-claude.sh --categories "clinical-biostatistics"

    The installer copies matching skills into ~/.claude/skills/ (default target). Use ./install-claude.sh --list to preview the skills first.

  • Claude Code / other agents — copy just this one skill: 
    cp -r bioSkills/clinical-biostatistics/multiplicity-graphical ~/.claude/skills/

    (run from inside your clone — the previous step left you in bioSkills/; otherwise replace bioSkills/ with the absolute path of your clone). Install the R packages on first use, e.g. install.packages(c("gMCP", "graphicalMCP", "multcomp")); the Python route needs pip install "statsmodels>=0.14" for multipletests. gMCP 0.8.16+ and graphicalMCP 0.2+ are the versions the skill targets.

What it does

Walks a decision tree from the trial’s claim structure to a specific procedure:

  1. Confirmatory or exploratory? A regulatory claim requires FWER control. Exploratory work can use FDR (Benjamini–Hochberg) — the skill treats mixing the two up as the primary error.
  2. Is there a strict clinical priority ordering? If yes, fixed-sequence (serial gatekeeping) or a chain graph.
  3. How many endpoints, and what dependence?
    • ≤ 3 endpoints, dependence unknown → Holm (universally valid)
    • ≤ 3 endpoints, positive correlation established → Hochberg or Hommel, after verifying PRDS
    • ≥ 4 endpoints with a primary / key-secondary / subgroup hierarchy → graphical procedure via gMCP
    • Multiple distinct families → parallel or serial gatekeeping
  4. Can endpoints move in opposite directions? (LDL vs HDL, efficacy vs safety) → Holm only — PRDS fails and Hochberg is anti-conservative there.

Underlying theory the skill implements: the closed-testing principle (Marcus–Peritz–Gabriel, with Goeman 2021 admissibility), Bretz–Maurer–Hommel graphical procedures, and gatekeeping in parallel, serial and mixed forms.

Calibration numbers it carries:

Threshold Value Source
Bonferroni power loss at ~10 tests under positive dependence 30–50% Sarkar 1998 (PRDS)
Hommel gain over Hochberg under PRDS 1–3% Hommel 1988, Biometrika
Subgroup α budget ≤ 20% of total Dane 2019, EFSPI white paper
Confirmatory FWER level α = 0.05 (two-sided) Convention

Primary use cases: allocating α across multiple primary endpoints, designing a key-secondary testing hierarchy, choosing between Holm and Hochberg.

Notes

Hochberg’s validity is conditional, and the check is the point. Hochberg and Hommel require positive regression dependence on subsets (PRDS); when it does not hold the Type-I error is inflated, and the skill’s rule is to revert to Holm (Sarkar 2008, Ann Stat). The opposite-direction endpoint case is the standard counterexample.

Regulatory anchor: the FDA Multiple Endpoints in Clinical Trials final guidance (October 2022). The graph — nodes, weights and α-propagation edges — belongs in the SAP before unblinding; a graph fitted after seeing the data controls nothing.

FWER and FDR are not interchangeable knobs. FDR is the right target for a screening analysis with many hypotheses; it does not support a regulatory claim on a single endpoint.

Distributed as a SKILL.md (plus reference material) in the bioSkills collection — Claude executes the workflow locally rather than as an MCP server. The upstream skill front-matter name is bio-clinical-biostatistics-multiplicity-graphical; if invoked as a namespaced plugin command it resolves under the bioSkills plugin, not as a bare /multiplicity-graphical. Upstream directory: clinical-biostatistics/multiplicity-graphical.

Related: Subgroup Analysis (which spends the ≤ 20% subgroup α budget defined here), Power and Sample Size (a split α changes n), Adaptive Designs (interim looks consume α through a spending function, on top of the endpoint split), Trial Reporting.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.