Check the structure of your gaze data¶
Before choosing a method, examine a long-format gaze CSV in your own browser. The inspector flags structural issues such as duplicated timestamps, nonnumeric coordinates, missing observations and repeated trials. It does not upload data, convert timestamps, infer missingness mechanisms, run the package, or claim valid statistical uncertainty.
Files are read only in your browser. Nothing is transmitted or uploaded. The exported JSON contains counts and diagnostics, not raw gaze rows. Maximum: 12,000 data rows and 1.2 million CSV characters. A subset check does not qualify an entire recording.
Choose a file, paste a CSV, or load the invented example.
Accepted columns and assumptions¶
Required literal headers: participant_id, trial_id, time_s, x, y. A condition column may be retained. Coordinates may be absent as empty CSV cells, never interpreted as zero. Time and coordinate units must be declared explicitly; no conversion is performed.
A consistent grid is only structurally compatible with dense analysis, not proof that the curve is scientifically meaningful. Native unequal support may require sparse methods. Coordinate-specific missingness is not, by itself, evidence of a device's asynchronous acquisition. Repeated trials must retain participant clustering.
Structural readiness is not method or inference validation
This inspector does not assess equipment calibration, drift, fixation/saccade classification, AOI labels, missing-data identification, statistical power or confidence-band calibration. Research issues #289 and #296 remain open.
Next: Input schemas · Method selector · Run synthetic starter