All binary missing-choice assignments give 10 to 12 Holm p-values less than 0.05 for the ds000005 loss/gain tests.

For the 16 combined-participant loss/gain tests in OpenNeuro ds000005, the recorded-choice calculation gives 11 Holm-adjusted p-values below 0.05. Across binary assignments to all 154 missing choices, the number below 0.05 is bounded by 10 to 12; saved assignments attain both limits. The analysis exhaustively enumerates the 1,051,492 assignments for 15 participants with at most 20 omissions, groups equivalent sufficient-statistic vectors, and uses valid Holm bounds for the participant with 82 omissions. These results apply to the pinned source files and fixed logistic model, with a common intercept across blocks, independent binary trials, and nominal chi-square likelihood-ratio p-values. The completion range is not a probability or confidence interval, and the coefficient ranges are not confidence intervals. The analysis does not establish parameter recovery, family-wise error control under unknown missingness, brain-signal results, or biological causality.

missing data · binary outcomes · sensitivity analysis · Holm correction · loss aversion · mixed gambles · logistic regression · OpenNeuro ds000005 · likelihood-ratio test