A count and trim rule with separate cutoffs matched 49 of 67 PVT score sets in ds004902 version 1.0.8

We tested public psychomotor vigilance task (PVT) scores in OpenNeuro ds004902, version 1.0.8. The 67 response files contain 3430 responses. Eight reported score sets matched full-file lapse count, median, and sample standard deviation under a common 500 ms lapse rule. A count and trim rule matched 49 sets to their own source files at 0.011 ms tolerance. This rule counted responses above 500 ms in normal sleep and 700 ms after sleep deprivation. It then used the other responses for median and sample standard deviation. The common compatible integer cutoff ranges were 499 to 501 ms for the 19 matching normal sleep files and 700 to 707 ms for the 30 matching sleep deprivation files. These ranges do not overlap. In 29 complete stored pairs, the common 500 ms rule gave a mean lapse fraction change of 7.92 percentage points. The separate cutoff rule gave -1.18 points. The paired difference was 9.10 points, with a 95% person bootstrap interval of 6.87 to 11.40. Fifteen score sets matched one different stored file. Article test values also matched different metadata columns from the stated labels. The package preserves source files, provenance, code, all results, and independent checks. Numerical matches do not prove participant identity or the original score code. The cutoff contrast is a score rule comparison under fixed source labels, rather than a causal sleep effect.

psychomotor vigilance task · sleep deprivation · response time · score audit · OpenNeuro · neuroscience data