Separate block intercepts change one of 16 adjusted ratio test decisions in ds000005.

We repeated the reported logistic analyses of the public OpenNeuro ds000005 mixed-gamble choices and tested a separate-intercept model on the same choices. Across 16 participants, the number of Holm-adjusted loss/gain ratio tests below 0.05 changed from 11 to 12; only sub-07 changed its decision. For sub-07, the adjusted p value changed from 0.18454198467168326 with one intercept to 0.03676829930019113 with three block intercepts. Both pointwise 95% profile intervals excluded 1. All 64 sample classifications, 236 likelihood-ratio checks, and 16 pooled profile intervals and adjusted p values matched the computational reference. The data comprise 3,942 recorded choices from 16 participants; 154 missing choices were omitted. This comparison uses the same cohort and choices, assumes independent choices and the stated logistic models, and does not establish that either model is correct or identify learning, fatigue, or biological causality. The versioned dataset DOI could not be confirmed for the pinned source; its Git commit and file hashes are recorded in the package.

mixed-gamble task · loss aversion · logistic regression · block intercepts · profile likelihood · OpenNeuro · ds000005 · reproduction