Deleting one recorded stop trial can shift omission-handling SSRT contrasts in two public tasks

Across two public stop-signal task datasets, deleting one recorded stop trial could shift the paired integration-SSRT contrast between maximum-RT replacement and omission deletion, even when go omissions were uncommon. In fixed baseline cohorts of 159 PIOP2 participants and 82 ds004636 participants, equal-participant mean contrasts were 11.453 ms and 1.817 ms. The largest absolute changes after any one-stop deletion were 85 ms and 67 ms; 36/159 and 6/82 participants had a maximum shift of at least 10 ms. Exact inverse-ECDF ranks show that the contrast is the sum of adjacent observed go-RT gaps between selected ranks, and that deleting a successful or failed stop produces two distinct probability changes. A <=5% omission screen reduced the cohorts to 152 and 77 while the largest shifts remained 85 ms and 67 ms. Linear interpolation gave different magnitudes, including largest shifts of 38.190 ms and 36.207 ms. The exhaustive deletion ranges are deterministic sensitivity of recorded estimators with go observations fixed; they are not confidence intervals, test-retest reliability, latent stopping ability, or a replay of the adaptive staircase. Cohort differences do not support biological comparison, and the results do not establish that omission deletion is preferable.

stop-signal task · stop-signal reaction time · omission handling · integration method · single-trial deletion · sensitivity analysis · inverse empirical CDF · reaction time