Articles¶
These articles explain the scientific decisions that sit between running code and making a claim. They complement the formal method pages and API reference.
Research workflow¶
From raw gaze to a defensible analysis¶
Why source identity, units, sampling cadence, QC, review, denominators, events, AOIs, grouping, and provenance need to remain separate stages.
QC flags are not exclusions¶
Why an anomaly score or trial-quality flag is evidence for review, not permission to silently remove observations.
AI proposals are not ground truth¶
How to use detectors, trackers, classifiers, and semantic models without allowing predictions to silently become empirical labels.
Missing is not zero¶
Why missing observations, absent-by-design states, undefined quantities, observed zeros, and censored latency outcomes require different representations.
Participant-held-out validation¶
Why the held-out unit must match the population to which an event-model claim is intended to generalise.
From output to reviewer-ready evidence¶
How to turn outputs into a reproducible evidence bundle without upgrading the scientific claim.
Continue deeper¶
Use the Method chooser for a decision tree, the Workflow gallery for complete pipelines, and the Plot gallery for visual diagnostics.