FRP and Deconvolution Use

Preserve canonical fields

For models that accept string or event-level predictors, EEG.event.type may be descriptive or restored to the original marker. Keep eyesort_full_code unchanged as the stable EyeSort identity, and retain original_type for traceability.

Useful predictors include current_region, region_pass_number, fixation_in_pass, regression flags, previous/next region fields, condition/item/trial numbers, word text/index, and imported IA covariates.

Model the scientific problem

EyeSort labels behavior; it does not deconvolve temporally overlapping brain responses. When event classes differ in fixation duration, saccade amplitude, timing, or participant/item prevalence, choose a model that handles those differences.

Reporting checklist

Report:

• EyeSort version and upstream synchronization/event-detection method;
• exact region geometry and label criteria, including saccade thresholds when used;
• event-marker format and conflict policy;
• post-EyeSort event/epoch counts by condition and behavior;
• participant- and item-level distributions of target behaviors;
• shared text-IA and label-queue files;
• downstream overlap, artifact, epoching, and statistical methods.

Explain the behavioral-specificity versus power tradeoff. A narrow label can be theoretically precise while leaving too few or unbalanced observations for reliable inference.