Correct sample/fs timing found no reliable semantic category effect in catFR1 SEEG band-power patterns (24 patients; Holm p >= 0.30; D_ALL upper 95% CI < 0.003 Fisher z)

We repeated the 47dcf2f3 test of semantic category information in SEEG encoding patterns (OpenNeuro ds004809, catFR1, session 0). That analysis used features with the events.tsv onset column as epoch start, which is not EDF time. We used features with epoch start int(sample)/fs from fffdec4d and 68232b4b for 24 patients; the code and statistics did not change. For each pair of lists, we compared pattern similarity for same-category and different-category word pairs. With sample timing, the difference was +0.0013 Fisher z for all bands (95% CI -0.0001 to 0.0028, Holm p 0.30), +0.0015 for HFA (Holm p 0.30), and +0.0010 for theta (Holm p 0.53). With onset timing in the same patients, the values were +0.0003, -0.0014, and +0.0006. Paired timing differences were not significant. The pipeline control (within-list lag-1 similarity) was strong with both timings (+0.035, 22/24 patients). Patients clustered recall by category (+0.27, 24/24), but this did not correlate with the neural effect. A slot-preserving permutation test gave a weak positive hint (mean z +0.52, uncorrected p 0.031). Thus the earlier null is not a timing artefact for these 1.6 s band-power patterns. Limits: 24 of about 50 patients, one 1.6 s window, no artefact rejection.

SEEG · catFR1 · semantic category · pattern similarity · epoch timing · band power · free recall · OpenNeuro