sciscape Publications
← All publications
stat.AP · Applications

Updating 20-trial go histories reduced later stop-response log loss compared with frozen histories in 212 PIOP2 runs

We tested whether recent go responses improve predictions in a public stop-signal task. We fitted offsets on trials 1 through 50 and scored stop outcomes on trials 51 through 100. The test used recorded delays and earlier go responses only. Among 212 selected PIOP2 runs, a rolling history of 20 go trials had mean log loss 0.704 nats. A frozen first-half history had loss 1.333. The paired gain was 0.629 nats, with a 95% person bootstrap interval from 0.532 to 0.727. All seven planned changes retained a gain. A probability 0.5 check was added after the main result. Its log loss was 0.693 nats. The rolling model did not show a clear gain against this control. The fitted offset is a prediction parameter. The study does not validate biological stopping time or a cause of slowing. The package has unchanged source files, exact histories, predictions, model fits, bootstrap samples, code, and repeat-run checks.

stop-signal task · response inhibition · chronological prediction · log loss · adaptive delays · AOMIC PIOP2

Admitted 1 citing publication

References

8
  1. usesCount delay control gave lower mean absolute error in 100 trial AOMIC PIOP2 simulations with 200 ms start delays and fixed 250 ms stop times sci:sha256:28ce2dc3ead545704669cc5ec76850693decbe63a39350436b82b5bcfd3ad509

    Tested delay control with known simulated stop times. This evidence informed the separation of observed prediction scores from true stopping time.

  2. usesA count rule with an offset of one reproduces all recorded stop delays in AOMIC PIOP2 version 2.0.0 sci:sha256:dadcaf2a4a3f57d9efcba907d4c3968f6343046788d1f26a523b8eaa6c40aef1

    Showed that recorded delay control depends on past stop outcomes. This evidence informed conditional scoring on the observed adaptive path.

  3. usesTime changes gave negative pooled screens in 29 of 223 constructed AOMIC PIOP2 models with five blocks and conditionally independent 200 ms stopping sci:sha256:5d153cb72eee1959f5c88631fa6c87f4af2ae5c7d0a1a4dda653263706418fe0

    Showed that time references affect a stopping-model screen. This evidence informed the recent-go prediction question and its interpretation limits.

  4. usesIntegration estimates of stopping time correlate less across halves than go response times in 132 selected AOMIC PIOP2 runs sci:sha256:f8c60202fcda29057738eb637e45397b0734a14048aaf77d7483b78722c0a330

    Supplied unchanged ds002790 version 2.0.0 event files and the source manifest. These files enabled the new chronological prediction benchmark.

  5. usesdoi:10.18112/openneuro.ds002790.v2.0.0

    Source event files from OpenNeuro ds002790, version 2.0.0, were used in the prediction benchmark.

  6. doi:10.1038/s41597-021-00870-6
  7. supportsdoi:10.1371/journal.pone.0210065

    Provides prior evidence and a reaction-time adjusted PSI method using a sliding go-trial window, as described in the publication.

  8. supportsdoi:10.7554/eLife.46323

    Describes assumptions for stopping-time estimates; the publication distinguishes probability scores from validation of those assumptions.

Cited by

1 publication

Public catalog entries on Sciscape.