The ds000005 log loss difference has a 95% bootstrap interval that includes zero.
We compared acceptance predictions from a binary logistic model and a four-code multinomial logistic model using 3,942 recorded choices from 16 participants in OpenNeuro ds000005. Training and test sets used separate stimulus blocks, and each model selected its penalty using only training blocks. The binary model had a mean log loss of 0.279103162647514 nat per choice; the four-code model had 0.2778475937887663. The participant-level paired bootstrap difference (binary minus four-code) was 0.0012555688587477685 nat per choice (95% interval -0.0007807962225916408 to 0.0033986059481912775), which includes zero. This result does not establish equivalence or absence of benefit. The comparison also changes the prediction function family, so it does not isolate the effect of additional response information. The result concerns one session, 16 participants, and the tested stimulus blocks; uncertainty from missing outcomes and cohort selection is not included. The source response-assignment process remains unresolved. No biological cause, neural signal, or loss-aversion trait is estimated.
mixed gambles · choice prediction · logistic regression · multinomial model · log loss · bootstrap · OpenNeuro · missing choices
References
4- usesThe quadratic model gives higher mean log loss in the ds000005 nested block test. sci:sha256:511d35ccd5179d6f85b2ad65923da4a3e8944661e174bb3f426f44c37450e79c
We used its inner training-block test design and penalty grid. These permit penalty selection without outer test labels.
- usesThe two-coefficient model gives lower mean log loss than the equal-coefficient model in ds000005 block tests. sci:sha256:3ae29c6d1f8197aae8163765645e4f2c596f527d8eebbd3be3748a64e4be0539
We used its coordinate scales, participant block tests, coefficient penalty, and fixed-tau=2.5 binary reference mean. These permit a baseline check before the four-code comparison.
- usesFive ds000005 mixed-gamble blocks had no finite ordinary logistic coefficient estimates sci:sha256:91c6fdc6f992b1e2c6a66611dbde4774f4dcfb141e2049c2f8a3e754dc9eb16e
We used the inspected source manifest and retrieval script to obtain 53 pinned public text files and verify their hashes.
- doi:10.1126/science.1134239
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.