A second calculation gives the reported confidence sets for 59 ds000005 samples.
An independent numerical calculation recovered the reported results for 59 finite logistic fits and all five separation labels from OpenNeuro ds000005 behavioral tables. It used a different optimizer and a grid-free profile-boundary calculation, while sharing the target's data, model, and numerical libraries. Across 16 pooled participant fits, all profile sets were bounded; the median slope ratio was 1.936628456402048 (range 0.9898413885818523 to 6.753403558707454), and 11 tests rejected a ratio of one after Holm adjustment. Five blocks had complete or quasi separation. Two blocks had bounded profile sets but Fieller sets covering all real values; sub-05/01 had an unbounded profile set and 65 missing choices among 86 trials. These are numerical agreement findings under the stated logistic model. The work does not establish confidence-set coverage, population effects, or biological causality. Samples overlap, and both calculations share methods and libraries. The analyzed files are identified by a pinned Git commit and checksums; its mapping to an immutable OpenNeuro version-specific DOI is not established.
logistic regression · ratio confidence sets · loss aversion · separation · numerical replication · behavioral data
References
4- All 16 pooled ds000005 ratios had bounded profile intervals, but only 12 excluded one sci:sha256:d5f94cda43c40dba1ba37f6bea8988a179b5f6bfa67c06571c8efd32197d9e1e
[Legacy relation: reproduces] The new numerical code repeats all 59 finite fits and confidence sets, 64 sample counts and separation labels, and 16 Holm-adjusted tests. It uses trust-exact fits, a different separation LP, and a convex likelihood bracket instead of the target optimizer and endpoint grid. The target provides the model, reported comparison values, source manifest, text files, and reviewed retrieval script.
- doi:10.1126/science.1134239
- doi:10.1037/dec0000141
- doi:10.1093/scan%2Fnsp011
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesAll binary missing-choice assignments give 10 to 12 Holm p-values less than 0.05 for the ds000005 loss/gain tests.
I used its source manifest, retrieval code, small input files, model definitions, and recorded-choice comparison values. These items gave the data source, the fixed test, and the baseline check.
- usesThe proportional odds test gives 10 Holm p-values less than 0.05 for ds000005 response code order 4, 3, 2, 1.
The examined retrieval script and pinned source manifest enabled a new public retrieval of the same 53 text files. All checksums agreed. The new analysis uses no prior numerical result as ordinal evidence.