Analyze the trajectory, not just the fixation count.¶
eyetrajectoriespy models gaze as a time-indexed two-coordinate functional process. Work with dense, sparse, irregular and repeated-trial trajectories while keeping the assumptions behind analysis and uncertainty visible.
Begin your first analysis Find a topic Install stable 1.2.0 Find a Python integration Try method selector Research evidence
Conceptual synthetic schematic, not measured human gaze or a performance claim.
Full-site usability and migration status¶
The site-wide quality programme now evaluates rendered internal links and deep anchors, topic search discoverability, simulated mobile Lighthouse performance, browser accessibility and complete Zensical route inventory. The selected Zensical Modern design is evaluated through a strict publication gate, with MkDocs Material retained as an independent compatibility and rollback baseline. Automated checks do not replace the open manual screen-reader and keyboard review.
Site transparency and accessibility¶
The website quality and discoverability guide explains how documentation pages, canonical URLs, visual assets and browser accessibility are tested. The checks are reproducible but do not certify WCAG compliance or guarantee search visibility.
Cite and reproduce research¶
Using the software in a paper? Get the exact 1.2.0 citation and research provenance checklist. The website citation reflects CITATION.cff, distinguishes released APIs from repository previews, and does not invent a DOI.
Start with your data and research question¶
New to functional gaze analysis? Use the data-to-method guide to identify the independent sampling unit, time grid, estimand and uncertainty scope before choosing an estimator. The quick start provides seeded executable examples.
Illustrated methods — see the analysis before the API¶
Browse a 12-figure scientific storyboard linking deterministic FPCA, sparse, regression and nonlinear plots to reproducible methods and their explicit interpretation limits. Each figure comes from the existing documentation build; no new results are claimed.
Interactive research tools¶
New: Inspect the structure of a gaze CSV in your browser — read participant/trial support, missing coordinates and timestamps locally without sending data to a server. This is a structural preflight, not an instrument or inference validation.
- Find a suitable analysis route
Answer four questions about your time grid, estimand, sampling hierarchy and uncertainty. Receive an interpretable, non-prescriptive method starting point.
Interactive selector · Ten-route comparison
- Explore a synthetic gaze trajectory
Change retained sample support and switch between the spatial path and time-indexed coordinate. Descriptive spread is explicitly not a confidence interval.
Trajectory explorer · Runnable synthetic dataset
- Search figures and research evidence
Filter existing W3 figures and R76–R95 research records without promoting experimental findings to stable package claims.
Filter figures · Search evidence
- Check documentation provenance
Use stable 1.2.0 online docs with release-indexed notes and immutable tagged source references.
Connect your gaze data to the right analysis¶
- 1. Identify the source. Browse 14 Python integrations by parser, data standard, device or functional-library role.
- 2. Verify observed support. Inspect a CSV locally or follow source metadata guidance without automatically filling missing samples.
- 3. Choose the estimand. Use the method selector or scientific data-to-method guide.
- 4. Audit the research claim. Check the qualification evidence before treating plots or fitted uncertainty as validated.
Python ecosystem interoperability¶
Bring preprocessed sample tables from upstream acquisition packages into the existing functional-data contracts without pretending that event detection validates functional inference. A new optional pandas/Polars/Parquet example shows strict source mapping and refuses missing or unequal dense grids.
Python integrations pymovements example BIDS metadata inspector Runnable table handoff
EyeLink acquisition interoperability¶
A new SyeLink 2.0.0 EyeLink ASC integration example maps explicitly selected-eye screen gaze samples without losing recording-segment boundaries. Explore the Python EyeLink parser comparison before choosing a source reader. Both are research previews rather than stable vendor-reader APIs.
Begin with your scientific question¶
- Explore variation in continuous gaze paths
Use FPCA and planar MFPCA to inspect structured temporal variation.
- Analyze sparse or asynchronous recordings
Respect irregular sampling and coordinate-specific recorded times.
Sparse planar methods · Irregular data
- Examine changing experimental effects
Fit interpretable functional regression with explicitly declared support and units.
- Model participants and repeated trials
Keep cluster structure and repeated observations visible.
- Study recurrence and temporal dynamics
Report embedding, threshold and time-delay decisions.
- Report auditable decisions
Surface preprocessing, model assumptions, scientific contracts and failures.
Released software is not the same as ongoing research¶
Available via pip
Stable version 1.2.0¶
The published package includes qualified 1.1 sparse/irregular methods and ten supported W1–W4 workflow routes under eyetrajectoriespy.workflows.
pip install eyetrajectoriespy==1.2.0
Scientific limitations
Released workflows and experimental research¶
Ten W1–W4 workflow routes are published stable APIs. Twelve E1–E4 helpers remain experimental, and the W3 whole-function confidence-band and field-MNAR claims are not qualified (issues #289 / #296).
Explore scientific evidence¶
Read each figure in the context of its source. R82 shows that apparently high gaze retention can coexist with inadequate independent-unit support. R95 proves conditional sharp population bounds given externally justified return assumptions; it does not establish the true missing-gaze mechanism or calibrate nominal-90% confidence bands.
Scientific figure atlas Interpret workflow figures Study map R76–R95
Latest news¶
24 October 2026 · Stable 1.2.0 released
The GitHub release and PyPI distribution passed production publication and exact-version installation checks. Statistical limitations remain explicit.
23 October 2026 · R93–R95 research evidence
Exact observational equivalence, externally assumed completion floors and monotone-return sensitivity, with independently audited records.
Transparent research software¶
The package starts after gaze has a scientifically interpretable time and coordinate representation. It complements, rather than replaces, fixation/saccade detection, instrument-level QC, AOI/scanpath methods and upstream recording workflows.
Representations · Mathematical reference · Workflow guide · Contribute · GitHub