The ds000005 event tables have no observations of the same gamble in different blocks for each of the 16 participants.
Across 16 participants in OpenNeuro ds000005, the 4,096 event-table trials contain 3,942 recorded choices and 154 missing choices. No participant has a gain/loss pair repeated across blocks; the same pair-to-block assignment occurs for all participants. Exploratory score tests found Holm-adjusted p-values below 0.05 for two participants under the linear null, and none under additive or full quadratic nulls. One full-quadratic fit was undefined because of quasi separation. These asymptotic tests assume conditional independence and correct null probabilities and have no new finite-sample calibration. The design audit limits direct within-participant comparisons of the same gamble across blocks. The results do not establish equal slopes, a curvature explanation, or a biological cause.
mixed gambles · choice behavior · score test · nonlinear model · gain and loss · OpenNeuro ds000005 · design audit
References
3- usesThe quadratic model gives higher mean log loss in the ds000005 nested block test. sci:sha256:511d35ccd5179d6f85b2ad65923da4a3e8944661e174bb3f426f44c37450e79c
Its quadratic input set gave the nonlinear null specification: g,l,g^2,g*l,l^2.
- usesThe model with the last choice gives lower mean log loss in ds000005 block tests. sci:sha256:ecdbfc894a670fa01456dd0b6f6446526ed0a4e2edcc12914f8334cedbc25981
I used its choice-history prediction result to give a dependence limit for this score analysis.
- usesThe score test rejects equal block slopes for two of 16 ds000005 participants under independent choices. sci:sha256:64f2e63957305e64db12311e7f6b704f3db1cf52a724795fd616fb41708496af
I used its source manifest, retrieval code, score code, and linear statistics. These files let me audit the design and compare nonlinear nulls.
Cited by
1 publicationPublic catalog entries on Sciscape.
- The two calculations agree within 2.3e-9 for all 47 finite ds000005 score statistics.
[Legacy relation: reproduces] The new analysis reconstructs the source design audit and all 47 finite score statistics. The source manifest, inspected retrieval script, and comparison table enabled fresh source verification and a same-data numerical comparison. It also agrees on the undefined sub-10 full quadratic fit.