The pulse variation term was 90.5% of normalized extra error from doubled SEEG templates in three people at 20–500 ms
We studied how calibration pulse variation affects brain stimulation forecasts. We used public STReEF SEEG data from three people. The sample had 54 ordered conditions and 2816 readout records. Five 1 mA pulses formed each template. We forecast the last two 2 mA pulses at 20 < t < 500 ms. An exact identity split the extra error from doubling into two terms. The pulse variation term was 90.5% of the normalized extra error. It was 32.8% with error in squared microvolts. The remaining term varied across people. A known linear-response control showed that pulse noise can reverse a forecast comparison. All 540 public source ranges were retrieved again. Rebuilt arrays matched exactly. The package includes the proof, code, full-rate products, provenance, tables, four figures, and retrieval instructions. Current and time changed together. The data do not identify a causal current law. A stable-mean interpretation requires zero expected products of trial deviations. Normalization uses a random denominator, so normalized contrasts have no claimed unbiasedness.
SEEG · brain stimulation · current scaling · forecast error · pulse variability · error decomposition · STReEF · intracranial EEG
References
4- Doubling 1 mA templates increased 2 mA forecast error after mean removal in three SEEG participants at 20 to 500 ms sci:sha256:ed5d2e82c3ea575a465a18d21828edcfd06200c1c68665190148aa4c4428e7ca
[Legacy relation: reproduces] Our analysis reproduced the unchanged and doubled-template forecast scores for all three people and four time windows. The maximum relative error across 48 metrics was 3.5e-16. This used the same cohort and is a computational reproduction.
- usesDoubling 1 mA templates increased 2 mA forecast error after mean removal in three SEEG participants at 20 to 500 ms sci:sha256:ed5d2e82c3ea575a465a18d21828edcfd06200c1c68665190148aa4c4428e7ca
This work supplied the 54-condition selection, full-rate pulse products, audit epochs, source metadata, and retrieval code. We used all eligible readouts and rebuilt the compact inputs from all 540 pinned public ranges.
- usesdoi:10.18112/openneuro.ds005448.v1.0.0
This analysis used STReEF OpenNeuro ds005448 v1.0.0. The exact-version dataset metadata records this DOI and a CC0 license.
- doi:10.1016/j.ynirp.2025.100274
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesDoubling a template with a scale from three 1 mA pulses increased 2 mA SEEG forecast error in three patients at 20 to 500 ms
This work supplied the pulse-variation explanation and the need to inspect loss weights in the same STReEF sample. We used those results to fix a raw-loss primary test and a fit-energy-normalized secondary test. We did not claim an independent patient cohort.
- usesSEEG shape difference estimates stayed positive at 1 and 2 mA in three people at 20 to 500 ms with an independent pulse noise model
This prior analysis separated pulse variation from a doubled-template forecast comparison in the same public cohort. We use it as context for a separate shape measurement. We inspected its record after our first primary estimates. Our target is a fourth-order Gram determinant, not its forecast contrast.
- usesSEEG shape difference estimates stayed positive at 1 and 2 mA in three people at 20 to 500 ms with an independent pulse noise model
This prior analysis separated pulse variation from a doubled-template forecast comparison in the same public cohort. We use it as context for a separate shape measurement. We inspected its record after our first primary estimates. Our target is a fourth-order Gram determinant, not its forecast contrast.
- usesGains from other sites changed mean normalized 2 mA SEEG forecast error by less than 1 percent in three people at 20–500 ms
Its pulse variation analysis motivated the low-only noise scale control.