With exchangeable cell labels, a fixed test found no clear ventromedial quantal pattern in 40 mouse GPe records

We tested whether a reported ventromedial (VM) quantal response pattern separates whole recorded cells. The v1.0.0 striatopallidal release supplied 40 mouse GPe cells and 19,562 workbook events. Source cleaning retained 19,556 events. A fixed Gaussian kernel compared event distributions with equal cell weights. VM versus 31 other cells gave squared maximum mean discrepancy (MMD) -0.0001121 and p = 0.3776 under 9,999 whole-cell label permutations. A four-region test gave p = 0.8313. A score that excluded each cell label from its reference means gave AUC 0.4014 and permutation p = 0.6193. Planned width, log-feature, kinetics-only and count checks gave no clear separation. A source audit found that one saved MATLAB cell slot copied 368 consecutive feature rows from another slot. A second pair shared 33 rows. The workbook and saved input had different cell counts. We rebuilt the released 19,371-row standardized matrix from the saved input with maximum error below 2.4e-13. In an added mean-contrast check, event randomization gave p = 0.0001 and whole-cell randomization gave p = 0.1322 for the same observed statistic. The tests assume exchangeable cell labels. Mouse identifiers and matched axon reconstructions are absent. These results do not establish equal regional physiology or test the paper's pharmacology. The package preserves source files, a cell record audit, event tables, code, null draws, figures and repeat instructions.

striatopallidal synapse · globus pallidus · quantal currents · cell replication · kernel test · data provenance · structure and function