The test did not establish lower hyperbolic than exponential loss: 164 ds002843 first visits, 90-day split, linear utility, 2% lapse
The study tested whether hyperbolic fits predict long-delay choices better than exponential fits. It used 164 first visits in OpenNeuro ds002843 version 1.0.1. Fits used 9,074 choices at delays up to 90 days. Tests used 10,572 choices at delays from 91 to 180 days. The main models assumed linear money utility and a fixed 2% uniform lapse. The mean exponential loss minus hyperbolic loss was 0.029795 nats per choice. The 95% paired person-bootstrap interval was [-0.015438, 0.075183]. The test did not establish a model advantage. Only the utility exponent 0.5 secondary interval excluded zero. All six secondary checks are preserved without multiple-test adjustment. Source retrieval, separate calculations, and a complete repeat passed. The package preserves source files, code, every test prediction, resample indices, results, and reproduction instructions. This is evidence on one task's prediction range. It does not establish a brain mechanism or equivalence of the models.
time discounting · hyperbolic model · exponential model · choice prediction · model comparison · out-of-range prediction · OpenNeuro ds002843 · person bootstrap · intertemporal choice
References
4- usesChanging money utility exponents from 0.5 to 2 increased fitted hyperbolic rates by a geometric mean factor of 9.57 in 118 ds002843 v1.0.1 adults with a fixed 2 percent lapse sci:sha256:bc1ca996ace11abfeec87982d5ba8af9013949c6079f4c33224275884c76ff01
This publication supplied 1,175 unchanged source files, their manifest, and preparation and retrieval code. Its fit procedure and utility results informed the new fixed model checks.
- usessci:sha256:9fa0ce68b90344ed7c34fd1ac8f08275070168c2f4e7c41247e1e867cd9397e2 not known to Sciscape
Its report showed that the prior test between visits used mostly repeated offers. This result motivated the new test outside the fit delay range. The cohort is shared.
- usesdoi:10.18112/openneuro.ds002843.v1.0.1
Source behavioral files and metadata for the analyzed first-visit choices; the saved record identifies version 1.0.1 and CC0.
- doi:10.1523/JNEUROSCI.2832-16.2017
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesImmediate reward fits did not establish lower hyperbolic loss: 178 AgeRisk people, two delayed rewards, linear utility, 2% lapse
The earlier short-delay to long-delay test informed this distinct reward timing split and the fixed model checks. This study uses new code and a different cohort.
- usesThe test did not show lower log-value loss: 164 ds002843 first visits, 90-day split, hyperbolic discounting, linear utility, 2% lapse
This study uses the first-visit raw files, filtered source manifest, and MIT retrieval code from this archive. Its difference formulas and ten-start grid method informed the new implementation. The new difference-rule scores were checked against its saved 164-person main results.