Phase 5: ERSP precompute and IC clustering

[neuromechanist]

Goal

Build a STUDY across subjects, precompute ERSP, and cluster components.

Entrypoint

src/matlab/phase5_ersp.m.

Steps

• Build STUDY from Phase 4 epoched sets with condition tags
• std_precomp ERSP: analysis 0-500 ms, baseline -500 to -100 ms
• Frequency range 4-40 Hz (Nyquist=50 Hz at 100 Hz sampling)
• Choose wavelet cycles (fixed 3 vs linearly increasing 3-8) - lock in plan-mode pass, justify in research.md
• Component clustering via k-means on dipole + scalp topography + rest spectra (not on ERSP itself, to avoid circular bias)
• Start with k=10; refine
• Identify occipital and temporal clusters of interest
• Save STUDY with precomputed ERSP under derivatives/ersp/

Success Criteria

• ERSP precomputed for every included subject
• Clusters identified and labeled (occipital, temporal, other)
• Design matrix documents LLR as covariate for Phase 6

Parent

Part of epic #1. Blocked by Phase 4.