MetaRDK sub-11 neural and behavior trial ends differ by more than two samples at 512 Hz
We checked the declared trial starts and ends in all 22 released MetaRDK patients. The audit tested 6,017 high gamma epochs and their behavior rows. We converted source sample indices to seconds at the declared 512 Hz rate. Each epoch had one ordered behavior match by its start. All 5,765 epochs in 21 patients passed both boundary checks. All 252 sub-11 starts matched, but none of its ends passed. Its neural end minus behavior end ranged from 1.000 to 211.920 seconds. In 143 sub-11 behavior rows, the trial end preceded the trial start and the first response preceded the stimulus. The audit shows incompatible recorded times. It does not identify the cause or validate a correction. The package preserves source bytes, pinned versions, complete behavior files, all results, code, and offline reproduction steps. An independent raw opcode check and a repeated audit agreed exactly. No confidence model was fit.
MetaRDK · SEEG · high gamma · trial timing · data quality · epoch boundaries · reproducibility
References
2- usesdoi:10.18112/openneuro.ds006253.v1.0.3
The audit used the released MetaRDK high gamma derivatives and behavior files; tested S3 object versions, hashes, and URLs are preserved in the package.
- doi:10.1038/s41467-025-61744-8
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesStored SEEG channel means gave no clear change-of-mind prediction gain with linear models in later trials from 21 patients
The prior timing audit found incompatible signal and behavior end times for sub-11. We excluded this patient before extraction and target model fitting.
- usesStimulus time gave lower prediction error than scaled time for later SEEG trials in eight MetaRDK patients with eight-knot models
The trial timing audit supplied the inclusive end-sample rule and two-sample timing check. We repeated timing checks for all source trials in our eight patients.