Event-Related Decoding#
Two lessons covering tasks where labels come from events and BIDS annotations rather than continuous state. Difficulty 2; assumes the core workflow track.
Event-locked decoding is where BIDS metadata earns its keep: you select
trials by trial_type, align windows to a stimulus or response onset,
and decode contrasts (target vs standard, congruent vs incongruent).
The visual P300 oddball lesson is the canonical first event-related
decoding task; the auditory variant is staged as a contrast that holds
the paradigm fixed and changes only the modality. Sourced from
docs/tutorial_restructure_plan.md Category C (lines 410-425), with
BIDS event handling per Pernet et al. (2019).
What you will learn:
How to filter recordings to include only event-locked trials of a given type using BIDS
events.tsvannotations.How to construct event-aligned windows around stimulus onsets, with matched baselines.
How to train a P3 target-versus-standard classifier and report performance with the appropriate chance level.
How to compare two event-related paradigms (visual vs auditory oddball) holding the analysis pipeline constant.
Run the lessons in order:
plot_20_visual_p300_oddball.py– visual P3 target vs standard.plot_21_auditory_oddball.py– auditory oddball as a controlled contrast.
How does the auditory P300 differ from the visual P300 of plot_20?