An assumed 1.953 ms time error limit gave no positive approximate 95% bootstrap lower bound for 66 intervals in 38 ds006780 v1.0.0 TD records
We tested whether an earlier audiovisual interval result remained under small response time errors. We used the same 38 TD records from OpenNeuro ds006780 version 1.0.0. Each finite time could change separately by up to one 512 Hz sample, or 1.953125 ms. Missing outcomes and trial assignments stayed fixed. We proved a necessary race inequality with smaller AV intervals and larger single-condition intervals. The bound is exact for one interval under separate corrections. None of 66 tested intervals had a positive approximate simultaneous 95% bootstrap lower bound with one sample allowed. For (150,225] ms, the mean guard gap was 0.947 percentage points, with a band from -0.953 to 2.848 points. The zero-error band was 0.080 to 4.393 points. A quarter-sample allowance also removed the positive lower bound. The error limits are assumed, not measured. This is an exploratory follow-up in the same records. The result does not prove a whole-distribution race or a neural mechanism. The package preserves the proof, code, 140 source files, hashes, all interval results, corrections, bootstrap indices, and instructions. Independent exact counts passed. Fresh source downloads reproduced all 22 result files and both figures byte for byte.
audiovisual integration · race model · response time · timing error · bootstrap · OpenNeuro
References
4- usesAn exploratory 66-interval audit found a small race bound violation in 38 ds006780 TD records with equal context weights sci:sha256:e1da790a1b38c216b5758bc3729797fc0a06c4adfca0992a0ea4ca195166e44f
This study used the earlier 38-person cohort, 66 endpoints pairs, trial rules, source package, source hashes, and reference arrays. It reproduced the zero-error result and added a test with bounded response time errors.
- usesA test at 11 times did not reject the audiovisual race bound in 38 TD records from ds006780 v1.0.0 with a 100 ms cutoff sci:sha256:d6c712ff30c7f235566d2e208e2f6c8970a14576fa9d5521f3ee71a74dad3cab
The event parser and fixed source retrieval code originated in this publication. They came through the earlier interval package. This study used that code, preserved its source and license, and checked it with an independent event scan.
- usesdoi:10.18112/openneuro.ds006780.v1.0.0
The study analyzes event data and task metadata from OpenNeuro ds006780 version 1.0.0.
- supportsdoi:10.3758/APP.72.3.839
Cited for the specific point that response censoring can change race-test inference.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.