Selection of correct GO trials changes 35 of 102 SSRT estimates with the ceiling rank in ds000030

Stop-signal reaction time (SSRT) is a model estimate of response stopping time. This sensitivity analysis used 126 CONTROL scanner runs from OpenNeuro ds000030 version 1.0.0, including pilot sub-10159, and compared all GO trials, removal of choice errors, removal of omissions, and removal of both. A prespecified screen retained 102 runs. For the ceiling-rank calculation, correct-only selection changed 35 screened estimates: 2 decreased and 33 increased; the mean change was +0.830569960865 ms and the median was 0 ms. Linear interpolation changed 53 estimates, with a mean change of +0.434832399837 ms and a median of 0 ms. On the 125 runs with all four calculations defined, the mean changes were -1.602963039936 ms (ceiling rank) and -2.242467518987 ms (linear interpolation); one run had no positive GO response after error removal. The results show sensitivity to GO selection, quantile rule, and sample screen. They do not show that correct-only selection improves SSRT, establish race-model validity, or imply a biological or clinical difference. Each participant contributed one run with 32 STOP trials; no population confidence intervals were calculated.

stop-signal task · stop-signal reaction time · SSRT · GO trial selection · choice errors · omissions · quantile · sensitivity analysis · OpenNeuro · ds000030 · neuroscience · reproducibility