A sequential stop-signal timing check has at most a 5% false flag probability per stream under known random assignment bounds and a conditional response floor
The proof gives a stop-signal timing check with a false flag probability of at most 0.05 in one trial stream. The check permits response speed to change and delay to depend on past outcomes. It requires a verified lower bound on each conditional stop assignment probability, a delay known before each assignment, and the conditional response floor. A recorded stop response must be at least as probable as a go response before that delay. Exact rational checks, an independent scalar calculation, and a separate replay checked the method. Constructed trials used all 226 AOMIC PIOP2 go response profiles from OpenNeuro ds002790 version 2.0.0. Each condition used 22,600 simulation runs. At 1000 trials, the boundary null flagged 3.978% of runs. With a 2200 ms delay limit, the check flagged 68.394% of constructed 25% stop response deletion runs. Detection was low at 100 trials. An invalid assignment bound caused almost universal flags under a valid response floor. These are simulation results. The public human records do not establish all design facts needed for a calibrated test. The package preserves source data, hashes, proof, code, results, example trials, and reproduction instructions.
stop-signal task · sequential testing · response inhibition · conditional randomization · supermartingale · simulation
References
8- usesA count rule with an offset of one reproduces all recorded stop delays in AOMIC PIOP2 version 2.0.0 sci:sha256:dadcaf2a4a3f57d9efcba907d4c3968f6343046788d1f26a523b8eaa6c40aef1
The recorded count rule supplied the adaptive delay controller in the constructed trials. A new source read reproduced all 7481 recorded stop delays. The long-run controller and later delay limit check are described separately.
- usesSixty of 171 ds005754 v1.1.1 records failed a timing test under one fixed go law per record and written stop fields sci:sha256:172c27be1f22ac52d7a9c69f329158f779fc1a6a63d4e2f106324f839746571a
The earlier timing-floor construction supplied the response-before-signal null. The new score removes its fixed go law requirement by using the conditional assignment design. No new claim is made about that publication’s human rejection count.
- usesIntegration estimates of stopping time correlate less across halves than go response times in 132 selected AOMIC PIOP2 runs sci:sha256:f8c60202fcda29057738eb637e45397b0734a14048aaf77d7483b78722c0a330
This publication supplied the initial ds002790 version 2.0.0 task files and pinned source manifest. We retrieved all 233 files again from their original pinned URLs. All hashes matched. The full 226-run sample supplied simulation go response pools.
- usesAn audit with fixed bins and trial counts limits false rejection of independent stop-signal races with a stable go distribution sci:sha256:81091280ded41e4323fe96f71f76b3a4cac035f9390160d725be16481accf7d2
The stable conditional go law required by the fixed-bin method motivated this trial-level check. The new score uses a known random assignment bound instead of one stable go distribution.
- usesdoi:10.18112/openneuro.ds002790.v2.0.0
The 226 go response profiles used to construct the simulation trials came from AOMIC PIOP2, OpenNeuro ds002790 version 2.0.0.
- supportsdoi:10.1038/s41597-021-00870-6
This data paper supplies the task description, including 100 task trials, randomized trial order, and an average stop fraction of 33%.
- doi:10.7554/eLife.46323
- doi:10.1093/jrsssb%2Fqkad009
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.