Removing trials with hippocampal endpoint runs of at least 1 ms changed the mean 8 minus 4 letter theta contrast by 0.002 dB in 14 ds004752 v1.0.1 patients under the stated windows and linear model

We tested whether removal of trials with stored integer endpoint runs changes a hippocampal theta load estimate. Public SEEG in OpenNeuro ds004752 version 1.0.1 supplied 14 epilepsy patients and 62 sessions. The outcome was theta power change from fixation [0,1) to late maintenance [4,6) seconds. A linear model adjusted for session, trial position, and previous load. The 1 ms screen removed 98 of 2,561 eligible trials, all from patient 12. The equal-patient mean 8 minus 4 letter contrast changed by +0.001876 dB. Its descriptive patient bootstrap range was [0, +0.005628] dB under independent, exchangeable patients. Shorter thresholds and separate session trends changed the result. All six patient 12 source files passed fresh complete checks. An offline rebuild reproduced all tables and draws exactly. Code, features, raw samples, source versions, hashes, results, and instructions are included. This is a secondary analysis of cited inputs. An endpoint run does not prove amplifier clipping. The result applies to the stated data, windows, reference, and model.

intracranial EEG · hippocampus · theta power · verbal working memory · artifact screening · endpoint runs · sensitivity analysis · OpenNeuro