Mean late response rank agreement decreased from 0.564 to 0.452 after sample variance subtraction in five selected SEEG patients

We tested whether sample variance subtraction improves late electrical stimulation response map agreement. We used nine bipolar pairs and 322 edges in five patients from OpenNeuro ds004457 version 1.0.2. The fixed window was 50 to 250 milliseconds. Mean Spearman agreement decreased from 0.564 to 0.452 after correction. The patient mean difference was -0.112, with an exact two-sided sign test p value of 0.0625. This small sample does not establish a population effect. With independent random pulse signs, expected mean waveforms were zero. Conventional map agreement remained 0.275; corrected agreement was 0.031. The correction estimates squared mean response energy when trials have equal means and independent residuals. Shared noise and repeated artifacts can remain. The correction is an established method. The new contribution is its checked SEEG test and conditional control. The package provides code, calibrated epochs, source versions and hashes, all results, and repeat instructions. All 54 raw objects were freshly checked. All 25 saved analysis files were repeated byte for byte. This is a new analysis of the parent clinical sample.

stereoelectroencephalography · electrical stimulation · response energy · sample variance subtraction · rank agreement · conditional control