JASP Quality Guide

Open the guide

An interactive guide to choosing and preparing analyses in JASP, with 54 analysis guides and 47 practical example datasets. Videos and annotated JASP files are being added.

Maintainer: Henrik Godmann (@hrgodmann)

Development

Requires Python 3.9+ and Node.js 18+. No additional packages are needed.

node validate.mjs
python3 build.py
node validate-media.mjs
node validate-brand.mjs
node validate-interface.mjs

The build produces two versions from the same content and application code:

• dist/index.html is the online guide. Videos and JASP files have separate, content-hashed URLs under dist/media/, so opening the guide does not download every tutorial. The media validator keeps its initial HTML below 2 MB and checks that it contains no embedded video or JASP payloads.
• dist/JASP-Quality-Guide.html is the complete standalone copy. It embeds all available videos, JASP files, examples, and branding, and works without a server or network connection. It is intentionally much larger. The online guide's Save complete offline guide action downloads this file.

To preview the online guide, run python3 -m http.server --directory dist 8000 and open http://localhost:8000/. Open dist/JASP-Quality-Guide.html directly to check the offline version. CSV downloads, saved analysis plans, and printing are available in both versions. Generated files in dist/ are excluded from Git.

Both versions retain the original media bytes, captions, and JASP download filenames. The build removes obsolete generated content-hashed files from dist/media/; it leaves unrelated files and assets-source/ untouched.

Contributing

Suggestions and corrections are welcome through issues or pull requests. Guide text and examples live in the JSON files; videos and annotated JASP files are configured in media.json. Connect each listed asset to an analysis through its override or an effective default; validation rejects unused assets so a finished tutorial cannot silently remain hidden behind a placeholder. Run the checks above after making changes.

Publishing

Checks run automatically on pull requests and pushes to main. To publish, set Settings → Pages → Source to GitHub Actions, then manually run Actions → Validate and publish guide on main. The existing workflow publishes the complete dist/ directory, including the online media files and optional standalone download; no deployment setting needs to change. Publish these outputs together so every media URL and the offline download resolve to the matching build.

Colleague previews — 7 November 2026

Ten new narrated MP4/annotated JASP pairs are integrated using the existing on-demand media convention. Their guides disclose the authorized data adaptations and known module limitations. Original datasets remain separately available; the eight-run factorial worksheet is distinct from the completed 16-run example. These are initial, unlistened previews for colleague feedback.

The two Bayesian sampling previews require the isolated repaired development build from hrgodmann/jaspAcceptanceSampling, branch revision/mixture-state-space-model, revision 9d08972b6b10b8d04b9650ecce4b1560f811b629. Activate that build in the JASP process before recomputing their companions. A package version label alone does not identify the implementation.

The final three Survival pairs (survival_accelerated, survival_mixture, and survival_recurrent) are also integrated as colleague previews, bringing the local guide to 54 linked pairs. They use the original website examples without deleting or changing observations. The narrated clips are 74, 55 and 64 seconds respectively.

These three require the isolated development build of jasp-stats/jaspSurvival, revision d3550d881f302762b35050a4c86e037378ab9179, package version 0.98.2, validated in JASP 0.99.0 on Apple Silicon. Activate that exact build before recomputing; stable Survival 0.96.6 lacks the advanced workflows. Independent statistical calculations, native save/reopen/recomputation, paired-data/settings/output comparisons, complete media decoding and sampled cut inspection passed. Narration transcription was compared with the scripts, but actual listening, continuously watched playback and real-browser player testing remain unperformed. Method-name pronunciation needs colleague listening, particularly Weibull, Arrhenius and Crow–AMSAA.

Accelerated B10 predictions stay at the four tested temperatures; condition numbers map to temperatures in the adjacent table. The mixture companion retains convergence/start sensitivity and a one-component comparison without claiming physical causes. Recurrent analysis retains all 24 positive-count exposure intervals; its bands concern fitted means and intensities, not future-count prediction intervals. Its wide prediction tables require horizontal scrolling. Reusable source recordings, narration, build and numerical validation evidence are retained in the development workspace for feedback; none of these additions has been published.