Early validation selects zero SEEG correction for later confidence forecasts in four of six MetaRDK records
We tested whether eight SEEG high gamma channel means improve forecasts of later confidence reports. We used six public MetaRDK records from OpenNeuro ds006253 version 1.0.3. Each model used early trials for calibration and later trials for testing. The test had 537 trials. Early validation selected zero neural correction in four records. Mean task MSE was 0.029981. Mean task plus SEEG MSE was 0.029943. The mean gain was 0.00003785, or 0.126 percent of task MSE. A descriptive person bootstrap interval was [0, 0.00010155]. The exact person sign-flip p-value was 0.50. The constant calibration-mean forecast had lower mean MSE, 0.023030. Thus, the task comparator was weak. Excluding the source marker channel left all main forecasts unchanged. The result applies to this selected clinical sample, fixed linear models, and offline source processing. It does not establish absence of confidence information or live BCI use. Fresh checks verified all six full source hashes and every archived channel mean. The package preserves all inputs, code, models, later forecasts, figures, source versions, hashes, checks, and reproduction instructions.
confidence forecasting · SEEG · neural decoding · MetaRDK · model validation
References
4- usesEight high gamma SEEG means do not reduce mean forecast loss for later task errors in six MetaRDK records sci:sha256:7beb394715f39fed5df47828831fba10e67541970afc2c613e2757003b311742
We reused its exact six-record trial means, source confidence labels, task inputs, timing masks, source receipts, and downloader. We fitted new confidence regression models. We did not reuse its fitted error models.
- usesAdding eight SEEG signals to a task model increases offline confidence forecast error on later trials in five of six MetaRDK sessions sci:sha256:f153e0af57732bd1a666747e35778c555a9091f0c980875e17ea5b2460f566ee
This publication first extracted and preserved the six public high gamma trial means and their full source hash receipts. Our input arrays came through the later error forecast package.
- usesdoi:10.18112/openneuro.ds006253.v1.0.3
The analysis uses processed trial means and behavior from this exact OpenNeuro dataset version.
- doi:10.1038/s41467-025-61744-8
Cited by
1 publicationPublic catalog entries on Sciscape.
- Early validation selected zero SEEG correction in seven of 15 additional MetaRDK records
[Legacy relation: reproduces] We kept the numerical confidence method fixed and matched all 537 prior marker-excluded main forecasts exactly. We then applied the same chronological selection method to the 15 other compact records. The primary 15-record estimate is kept separate from the prior six records.