Start here¶
Find a scientific topic or research question
From continuous gaze data to a defensible functional analysis. Published eyetrajectoriespy==1.2.0 supports continuous, sparse and irregular methods, plus ten stable module-scoped W1–W4 orchestration routes. Separate E1–E4 helpers remain experimental.
- Install stable 1.2.0
pip install eyetrajectoriespy==1.2.0
- Understand your data
Choose common-grid, paired sparse, asynchronous coordinate-specific or repeated-trial representation before model fitting.
Data-to-method guide · Input schemas · Representation guide
- Choose a scientific method
FPCA/MFPCA, functional regression, repeated effects, generalized responses or recurrence.
- Learn through examples
Inspect deterministic synthetic illustrations and the assumptions behind every plot.
Citing the software¶
Copy the software citation and follow the reporting checklist. This is the documented MIT 1.2.0 software record, not a claim of calibrated W3 confidence bands or qualified vendor adapters.
Begin with a complete first analysis¶
Follow the first-analysis walkthrough · Browse the visual methods explorer
The walkthrough distinguishes recorded-data readiness from synthetic method illustrations, scientific uncertainty and publication claims.
Practical interactive tools¶
Local CSV structure inspector — check long-format columns, time support and repeated-trial structure before import. It does not impute gaze or diagnose a missingness mechanism.
Choose a starting method Compare ten workflows Run a small synthetic CSV
The selector is guidance, not automated model validation. The downloadable teaching fixture is not empirical gaze or calibrated uncertainty.
Connect other Python eye-tracking packages¶
Choose a source and build a research checklist
The integration hub compares pymovements, PyNeon, Polars, Pandera, eyekit, PupEyes, Eye-Tracking-BIDS and existing functional-data extras. Start with the optional, fail-closed table example only after recording true column mapping, time and coordinate units.
New preview examples: pymovements 0.28.0 Gaze-object handoff and read-only Eye-Tracking-BIDS sidecar inspection. These do not install as stable package importers.
EyeLink path: SyeLink 2.0 handoff · EyeLink ingestion options. Source parser compatibility is separate from qualified inference.
Three questions before any fit¶
- Observation unit — independent curve, participant, trial or repeated participant cluster?
- Time representation — aligned common grid or irregular/asynchronous native samples?
- Uncertainty target — descriptive variation, prediction, pointwise effect or simultaneous functional inference? Not all targets are qualified.
The scientific contracts document when an analysis refuses to proceed. The research evidence library documents important experimental failures rather than hiding them.
What is actually available?¶
| Surface | Today | Where to read |
|---|---|---|
| Stable package | Published 1.2.0, Python 3.11–3.13 | Stable release |
| Sparse/irregular 1.1 methods | Stable qualified module-scoped methods | 1.1 API |
Ten transparent run_*_workflow() routes |
Published stable 1.2.0 module API | Workflow guide |
| R76–R95 W3 inference studies | Separate research and evidence, not production estimators | Research studies |
Do not infer that a research figure corresponds to a released estimator or calibrated confidence interval.