Independent mark and exchangeable five-minute block references gave different half-hour STE HFO map decisions in 18 SEEG records

We compared the first two half hours of 18 public SEEG records from OpenNeuro ds005398 version 1.1.3. We used 23,297 released STE HFO marks on 1,285 good bipolar channels. Individual mark allocation rejected a common channel map in 12 records after Holm correction. Five-minute block allocation rejected it in zero records with the same score and correction. The event reference assumed conditional exchangeability of individual marks. The block reference assumed exchangeability of joint count vectors across blocks. A diagnostic with the block row-total distribution also gave 12 corrected decisions. Row-total variation alone did not explain the decision difference. The result shows that allocation assumptions affect evidence for half-hour map changes. It does not establish biological map changes. The package contains pinned public inputs, source code, results, figures, and a full offline repeat.

high frequency oscillations · stereo-EEG · allocation assumptions · block permutation · conditional exchangeability · channel maps · likelihood ratio · Hellinger distance