A fitted constant reduced later choice log loss in 107 selected NARPS adults with a slope penalty of 1
This study compared two logistic models of public NARPS mixed-gamble choices. Both models fit person gain and loss slopes. One model also fit a constant acceptance term. The primary test fit runs 1 and 2 and predicted runs 3 and 4 in 107 selected adults. Amounts were in units of 10 ILS. The fixed slope penalty was 1. The free constant reduced mean log loss by 0.006599 nats per choice. The 95% person bootstrap interval was 0.002631 to 0.011895. Weaker penalties and other run splits had gain intervals that included zero. Separate all-run fits gave a median factor of 1.179 between the two loss-to-gain slope ratios. The ratios were on opposite sides of 1 in 31 people, with one more estimate at the numerical boundary. The source sample was selected with whole-task behavior. The package preserves source files, fixed hashes, a local plan, code, all fits and predictions, bootstrap draws, and separate checks. A full repeat matched all 47 result files byte for byte. The results concern these fixed models and this selected sample. They do not establish true loss aversion or a neural mechanism.
mixed gambles · choice behavior · logistic regression · constant term · held-out prediction · loss-to-gain slope ratio · NARPS · OpenNeuro · person bootstrap · model sensitivity
References
4- usesPerson slopes reduced test log loss against a person bias model in 108 selected NARPS adults under fixed logistic models sci:sha256:23e1e5766dd32b526e64226ad487117bdf43115ac61fb8114fad718b903082df
The unchanged event tables, five source metadata files, fixed source manifest, and retrieval code enabled this new constant-term comparison. A fresh public retrieval confirmed all files. No prior model estimates or prior analysis code were used.
- usesdoi:10.18112/openneuro.ds001734.v1.0.5
This is the exact OpenNeuro dataset release used. The source files are pinned to its official Git commit and their hashes are recorded in the package.
- supportsdoi:10.1038/s41597-019-0113-7
This data paper documents the NARPS task, source sample, and the possible reversed response buttons for sub-056 used to define the primary exclusion.
- doi:10.7554/eLife.80979
This related study fits logistic constants and gain and loss slopes to the same source data. It provides context for the present fixed-penalty held-out comparison, which does not reproduce its estimates or neural model.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.