A logistic choice model gave bounded positive loss-aversion ratio ranges for all 108 selected NARPS records at cutoff 3.841

We studied the gain-loss coefficient ratio used to measure loss aversion in mixed-gamble choices. We used all four runs for all 108 selected people in OpenNeuro ds001734, version 1.0.5. The binary logistic model had a free intercept, no penalty, and no slope sign restriction. All records had bounded positive profile likelihood ranges at cutoff 3.841458820694124. This included two records with no finite full coefficient estimate. Their main ranges were [0.8826690580, 1.1173315890] and [0.8972256709, 1.1214180721]. At cutoff 1, the complete separation record had the exact open range (8/9, 10/9). Exact boundary limits explain why a continuous root solver can fail there. A separate parser and optimizer checked the fits. A full repeat from new public downloads gave identical main result files. The package preserves source files, hashes, prior certificates, new code, results, and repeat instructions. The result applies to this selected cohort and model. Confidence coverage was not tested.

profile likelihood · loss-aversion ratio · logistic regression · complete separation · quasi separation · mixed gambles · NARPS · ratio inference