Python-native guides¶
The frozen article corpus preserves the scientific workflow context of the R 0.11.1 reference. These guides provide the complementary Python-first route: start from a research problem and compose current eyeprocesspy functions into a complete workflow.
Trial-level multilevel mediation¶
Prepare repeated-measures gaze mediators without collapsing trials: within/between decomposition, missing-vs-zero semantics, quality audits, provenance, and downstream Bayesian handoff.
End-to-end eye-tracking¶
Import, validate, audit, derive process structure, visualize, and preserve provenance.
Event detector multiverse¶
Propagate detector uncertainty from events through AOI features to the same statistical model, with explicit failures, provenance, and robustness diagnostics.
Open guide · Visual atlas · Decision guide · Troubleshooting · Reporting template
Gazepoint import & QC¶
Vendor detection, gaze/fixation/event/biometric files, pairing, combined imports, and QC.
Pupillometry¶
Canonical pupil streams, baseline/missingness decisions, functional representations, and reporting.
Process quality & uncertainty¶
Reliability, Bland–Altman evidence, sampling irregularity, calibration error, and probabilistic AOIs.
Standardized data quality¶
Separate accuracy, RMS-S2S, population-SD precision, BCEA, realized sampling behavior, jitter, and data loss; preserve metric-specific units and review-only quality rules.
Censored gaze latency¶
Retain valid never-inspected trials as right-censored observations, choose Cox/AFT estimands explicitly, diagnose assumptions, and report sensitivity transparently.
AOI perturbation & uncertainty¶
Stress-test AOI geometry, assignment stability, feature recomputation, and model coefficients under defensible boundary changes.
Open guide · Analysis plan · Troubleshooting · Reporting bundle
Psychometrics & IRT¶
Information, fit, DIF, process-informed measurement, score uncertainty, and validation evidence.
Reproducibility & release evidence¶
Benchmarks, provenance, manifests, validation evidence, frozen-R checks, and distribution verification.
For compact copy-paste workflows, use Runnable examples. For exact source-port context, use the 88-article library. For visual discovery, use the plot gallery.
Bayesian networks¶
Use the Bayesian-network guide for provenance-aware downstream dependency modeling from validated eye-tracking, pupil, physiology, questionnaire, and behavioral features. The workflow covers the data contract, theory-constrained structure learning, bootstrap stability and participant-grouped validation, detector/AOI sensitivity, and reporting.
The BN layer is deliberately downstream of event detection, AOI assignment, pupil/physiology preprocessing, and feature construction; a learned directed edge is not, by itself, evidence of a causal effect.