iMast

iMast is a free, open-source, graphical statistical software for reference material research and clinical measurement statistics, built as an extension of jamovi. It implements 153 statistical methods across 15 self-developed R modules, covering reference-material characterization, commutability, homogeneity and stability, measurement uncertainty (GUM/JCGM 100), and the relevant Clinical and Laboratory Standards Institute (CLSI) evaluation protocols (EP05, EP06, EP09, EP10, EP14, EP15, EP17, EP21, EP24, EP30).

iMast is published in the Journal of Statistical Software; see Citation below.

Key features

• Reference-material-centered clinical statistics — commutability (EP14 / EP30 / IFCC), characterization and value assignment, between-unit homogeneity, long-term stability, measurement uncertainty (GUM/JCGM 100), plus supporting precision, linearity and method-comparison procedures.
• Graphical user interface — spreadsheet data editor, point-and-click analyses, exportable tables and plots, built on the mature jamovi GUI framework.
• Assumption checks before analysis — distributional and homogeneity tests are run automatically to guide method choice.
• AI-assisted workflow — an optional large-language-model assistant (DeepSeek) helps with data import, variable-type inference, and analysis guidance. Statistical computation remains fully deterministic in the R engine; the LLM only assists natural-language-to-workflow conversion.
• Docker-based deployment — zero-configuration installation on Windows via an Electron desktop wrapper that imports the Docker image.

Repository layout

imast/
├── modules/            # 15 jamovi R modules (DESCRIPTION + R/*.b.R, *.h.R)
│   ├── ReferenceMaterial/        # commutability, homogeneity, stability,
│   │                             #   characterization, uncertainty, equivalence
│   ├── CommutabilityAndStability/
│   ├── Appraisal/                # appraisal ratio studies
│   ├── ANOVA, Descriptive, Estimation, Regression, ChiSquare, ...
├── llm/                # Python AI service (FastAPI)
│   ├── server.py                 # LLM bridge, tool calling, skills
│   ├── page.html                 # chat UI
│   ├── skills/                   # preset reference-material analysis workflows (JSON)
│   └── workflows/                 # workflow parameter templates (JSON)
├── desktop/            # Electron desktop wrapper (main.js, preload.js, loading.html)
├── data/               # example clinical / reference-material datasets (CSV)
├── replication/        # standalone R replication script for the paper
└── LICENSE             # GNU AGPL v3

Installation

iMast runs as a Docker container that wraps the jamovi engine, launched by an Electron desktop application.

1. Install Docker Desktop and start it.
2. Install the iMast desktop application (Windows installer provided separately).
3. Launch iMast; on first run the Docker image jamovi/jamovi:2.7.2-imast is loaded and the container is started automatically.
4. The application opens at http://127.0.0.1:41337.

To run the LLM assistant, set your DeepSeek API key in the application Setup dialog (or via the DEEPSEEK_API_KEY environment variable). The AI feature is optional and is not required to run any statistical analysis.

Reproducing the paper's results

The numerical results and figures in the JSS article can be reproduced with base R and two CRAN packages only — iMast itself is not required:

install.packages(c("outliers", "mcr"), repos = "https://cloud.r-project.org/")
setwd("replication")
source("imast_replication.R")

This reproduces the three case studies (EP05 precision, EP09/EP14-A3 method comparison and commutability, GUM uncertainty). See replication/REPRODUCIBILITY.md for expected output values.

Citation

If you use iMast in your research, please cite the JSS article:

Z. Zhou, Y. Yin, and J. Chen. (year). iMast: Integrated Statistical Software for Reference Material Research with Assumption-Guarded Workflows and a Metrological Equivalence Model. Journal of Statistical Software, vol.(issue), pages. URL https://jstatsoft.org/...

(Bibliographic details are completed at publication.)

License

iMast is free software released under the GNU Affero General Public License v3.0 (AGPL-3.0), the same license as jamovi. See LICENSE for the full text. Some upstream jamovi components are GPL-2+, which is compatible with AGPL-3.0.

Authors

• Zhiwei Zhou — Peking University Cancer Hospital & Institute
• Yongfeng Yin — Beihang University
• Jinfeng Chen (corresponding) — Peking University Cancer Hospital & Institute, chenjinfengdoctor@bjmu.edu.cn