RepTrace exposes modules for metadata preparation, manifest validation, MNE time decoding, result aggregation, calibration reporting, plotting, inference, paired decoder statistics, probability-trace onset detection, continuous raw-stream stimulus scanning, stream-level stimulus event detection, onset chunk validation, multi-task onset workflows, onset sensitivity analysis, sticky temporal modeling, emission comparison, semantic-stage analysis, and the calibration-aware temporal-state workflow.
Key command-line modules include:
Reusable table-oriented APIs include:
reptrace.metrics for calibration/probabilistic scoring metrics, pre/post window comparisons, and confusion-table summaries.reptrace.continuous_stimulus_scan for training an event-locked decoder on one raw run, scanning a held-out raw run, exporting long-stream class probabilities, and scoring detected events.reptrace.stimulus_detection for detecting zero, one, or many stimulus events in long probability streams and evaluating them against annotation tables.reptrace.results for time-decoding aggregation, participant/window result tables, and peak-window diagnostics.