API and plotting reference¶
The eyeprocesspy public surface preserves the frozen eyeprocess 0.11.1 API ledger and also supports explicitly documented post-parity extensions. Trial-level multilevel mediation preparation is one such additive extension; it does not alter the frozen parity record.
Browse the generated API Plotting reference Gaze-survival API Detector multiverse API Measurement-validity API Visual gallery
Scientific families¶
| Family | Representative surface |
|---|---|
| Core data contracts | EyeDataset, schemas, canonical tables, provenance, coordinate spaces, timebase |
| Import and adapters | generic readers, adapter registry, folder workflows, vendor detection |
| Gazepoint | gaze/fixation/event/biometric readers, pairing, real-export validation, workflow helpers |
| Preprocessing and features | fixations, saccades, AOIs, dwell, scanpaths, transitions, entropy, wide/trial feature tables |
| Detector robustness | detector specifications, event matching/agreement, annotated-event benchmarking, parameter sensitivity, AOI/feature propagation, inference stability, reports |
| Measurement validity & dynamic gaze | spatial error fields, pupil-size artefact correction, dynamic AOIs, pursuits, microsaccades, naturalistic coordinates, scanpath sensitivity, gaze stress tests |
| Censored gaze latency | right-censoring preparation, Kaplan–Meier, clustered Cox, AFT, diagnostics, sensitivity, reporting |
| Pupil and multimodal | pupil preprocessing, functional pupil, missingness, synchronized/staged multimodal workflows |
| Trial-level mediation preparation | within/between decomposition, mediation-level audits, explicit missing/zero states, trial-count support, provenance, serial/moderator preparation |
| Process quality | reliability, Bland–Altman, sampling irregularity, calibration error, gaze precision, quality profiles |
| AOI uncertainty | calibration propagation, probabilistic AOIs, compositional AOIs, sensitivity/boundary uncertainty |
| Psychometrics and IRT | information, scoring, fit, Q3, DIF/DTF, dynamic/process-informed/advanced IRT |
| Measurement intelligence | measure registry, guardrails, linking, norms, fairness, item-bank optimization |
| Validation | recovery, SBC-style evidence, stress tests, negative controls, grouped validation, atlases |
| Reproducibility | benchmarks, provenance manifests, software-paper evidence, parity/release validation |
| Plots and reporting | Matplotlib graphics, diagnostics, evidence visualizations, publication-facing reporting helpers |
Plotting surface¶
The plotting reference documents the major Matplotlib families and how to save, combine, and audit figures. Core plots return standard Matplotlib axes and commonly expose the underlying plotted data through ax.eyeprocess_plot_data; matrix plots can also expose ax.eyeprocess_plot_matrix.
How to find the right function¶
- Start with Getting started for the canonical workflow.
- Use Python-native guides when you know the research problem but not the function names.
- Use Runnable examples for compact scripts.
- Use the 88-article library for full workflow/parity context.
- Use Parity and validation for scientific fidelity and documented R/Python differences.
Reference breadth¶
The frozen parity surface contains 1,182 resolved APIs spanning import, data contracts, preprocessing, gaze/AOI analysis, pupil workflows, process measurement, IRT, validation, reproducibility, interoperability, governance, storage, scientific plots, and reporting. The mediation preparation module adds 10 public functions and 1 public data-contract class beyond that frozen ledger.
The generated API page is intentionally comprehensive; the curated guides and plotting reference provide the higher-level navigation layer.
Bayesian-network API¶
The Python-native Bayesian-network extension adds explicit node/data/constraint contracts, pgmpy-backed static structure learning and parameter fitting, posterior queries, participant-level bootstrap stability, grouped validation, detector/AOI sensitivity, CPT and structural perturbation sensitivity, discretization/noise/sample-size robustness, predictive calibration, plots, and reproducibility-oriented reporting. It is additive and does not alter the frozen R parity ledger.