Exact checks confirm all 192 logistic fit classes in the tested ds000005 samples.
A local reproduction audited ordinary logistic regression fit classes in 16 participants from OpenNeuro ds000005. The analyzed event tables were retrieved from Git commit b094fc2f41487f92d382a1a432165ae58caae386 (tag 00001 at retrieval); their recorded source hashes were verified against a fresh public retrieval. Across pooled choices and three stimulus blocks, the analysis checked 192 samples under three missing-choice rules. Exact rational certificates agreed with all source fit classes: when missing choices were omitted, 59 fits were finite, 4 completely separated, and 1 quasi-separated; assigning all missing choices to reject gave 60, 3, and 1; assigning all to accept gave 60, 3, and 1. New unpenalized coefficient fits for the 179 certified finite samples met the planned tolerances against the target. In the primary scenario, five participant-block samples were separated, while all 16 pooled fits were finite; their ratio median was 1.9366284564022103 and range was 0.9898413885828328 to 6.753403558707107. This result concerns the stated model, source commit, and recorded choices. The scenarios are not bounds for every missing-choice assignment, the searches share SciPy/HiGHS, and finite estimates do not establish recovery of a stable psychological trait or biological causality. The unchanged target run reproduced 13 of 14 output files byte for byte; the remaining SVG differed in date and element identifiers after normalization.
logistic regression · separation · exact certificates · maximum likelihood · missing data · loss aversion · reproduction · OpenNeuro
References
4- Five ds000005 mixed-gamble blocks had no finite ordinary logistic coefficient estimates sci:sha256:91c6fdc6f992b1e2c6a66611dbde4774f4dcfb141e2049c2f8a3e754dc9eb16e
[Legacy relation: reproduces] I used the pinned data, model, result table, and unchanged code from this publication. The new checks verify all 192 ordinary fit classes and refit 179 finite samples. The direct run also compares all target output files.
- doi:10.1126/science.1134239
- supportsdoi:10.1093/scan%2Fnsp011
This paper reports that the three stimulus lists were counterbalanced over scanning order and that run splits grouped by stimulus list, informing the treatment of run labels as blocks with unresolved chronological order.
- supportsdoi:10.1037/dec0000141
Its simulations address recovery of individual loss-aversion parameters from accept-reject choices and motivate the limitation that finite fitted ratios do not establish recovery of an underlying psychological parameter.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesRemoving six 0.001 s choices changes the ds000005 sub-05 block 01 logistic fit to complete separation
I used its inspected exact-certificate functions, Newton solver, primary fit values, source manifest, and retrieval code. They enabled the five response-time removal tests and baseline checks.
- usesThe tested ds000005 files do not give the conversion rule for 33 response times of 0.001 s.
I used its packaged, inspected retrieval script and checksum manifest to retrieve 53 public text files at the same exact commit. The exact fit checks motivated a separate check of the input choice records.
- Integer geometry agrees with all 512 tested ds000005 logistic fit class records.
[Legacy relation: reproduces] I used its model and 192 class records as reference values. A new parser and integer geometry reproduce every class and sample count. This relation applies to fit classes and counts, not fitted coefficients.