sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

Updating GO histories reduced later log loss against frozen histories in 266 UCLA scanner runs under fixed rules

A fixed recent-GO rule was tested in an independent public cohort. In 266 UCLA scanner runs, rolling histories reduced later STOP response log loss against frozen histories by 0.022177 nats. The adjusted 97.5% person bootstrap interval was 0.001219 to 0.044091 nats. The gain against a fixed probability of 0.5 was 0.017835 nats. Its interval was -0.003695 to 0.038243 nats and included zero. Each model fitted a personal offset on early trials. Numerical rule settings came from an earlier AOMIC study. The test used planned current delays and strictly earlier GO records. Seven setting checks and a check of 267 training runs showed differences across rule and task settings. The package preserves source files, hashes, a local plan, code, forecasts, fits, bootstrap samples, and independent checks. A full repeat with fresh source downloads matched all 48 analysis outputs. The result concerns conditional probability prediction. It does not estimate biological stopping time.

stop-signal task · response prediction · log loss · rolling history · GO response time · UCLA CNP

Admitted 0 citing publications

References

8
  1. supportsUpdating 20-trial go histories reduced later stop-response log loss compared with frozen histories in 212 PIOP2 runs sci:sha256:c1a30c9ecefea4fb86edd08fd281db32ee8bd203187c0d458a814b552a0a1b64

    The independent UCLA scanner test found a smaller gain in the same direction against frozen histories: 0.022177 nats, with an adjusted 97.5% person bootstrap interval of 0.001219 to 0.044091 nats. The gain against probability 0.5 remained uncertain. This supports the comparator-specific prediction result under the fixed tested rule.

  2. usesUpdating 20-trial go histories reduced later stop-response log loss compared with frozen histories in 212 PIOP2 runs sci:sha256:c1a30c9ecefea4fb86edd08fd281db32ee8bd203187c0d458a814b552a0a1b64

    This publication supplied the fixed probability rule, 20-trial history length, bandwidth, mixture, offset grid, early fitting split, and comparator design. We fitted new personal offsets on early UCLA trials and tested later trials in an independent cohort.

  3. usessci:sha256:dd489e042443f50582cf0f0cf8903b0df11d61e5c9b4a65f4b03eb6b5322eff4 not known to Sciscape

    This audit identified outcome-related zero values in the recorded StopSignalDelay field. Its field and unit checks informed our use of LadderTime as the planned current delay. A new sidecar audit verified the planned field as the pre-outcome staircase state.

  4. usesDiagnosis terms did not show a clear scanner stop response prediction gain in 132 ds000030 v1.0.0 adults with fixed models fitted on separate people sci:sha256:ca715176d619f08bd1272494cf8acaf6686464e6e2bcadcbc378b82deaabb359

    This archive supplied 1072 unchanged source files and pinned source records for OpenNeuro ds000030 version 1.0.0. Fresh network downloads matched all source hashes. We read the earlier parser to check field meanings. We did not execute its analysis code.

  5. usesdoi:10.18112/openneuro.ds000030.v1.0.0

    The publication analyzes the scanner and training task event files from OpenNeuro ds000030 version 1.0.0.

  6. supportsdoi:10.1038/sdata.2016.110

    Describes the UCLA Consortium for Neuropsychiatric Phenomics cohort and its task data.

  7. doi:10.1371/journal.pone.0210065
  8. doi:10.7554/eLife.46323

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.