drchronx/ai-agent-research-starter-kit/bci-results-reporting

BCI Results Reporting

Use this skill to turn EEG/BCI modeling outputs into manuscript-ready results.

Inputs

• Model outputs and metrics.
• Split design and leakage audit.
• Baselines and comparison models.
• Target journal or conference style.

Workflow

1. Report the data split first, because it defines the claim.
2. Provide subject-wise and aggregate metrics.
3. Include baseline and chance-level comparison.
4. Report uncertainty: confidence intervals, standard deviation across subjects, or paired tests.
5. Include confusion matrix and class-level metrics for imbalanced tasks.
6. Disclose preprocessing, feature extraction, tuning, and leakage safeguards.
7. State deployment limits: within-subject, cross-session, cross-subject, offline, pseudo-online, or online.

Hard Constraints

• Do not report only the best fold or best seed.
• Do not omit failed subjects.
• Do not describe offline results as real-time BCI.
• Do not hide leakage risks or post hoc model selection.

Output

Results table plan:
Subject-wise table:
Model-comparison table:
Figure plan:
Leakage and split disclosure:
Results paragraph:
Limitations paragraph: