SPK omission vs stimulus timing
ΔT = Tom − Tstim, for the 139 resolved units that also have an analysable stimulus arm.
descriptive 104/139 = 74.8% of units have ΔT > 0 (Clopper–Pearson 95% CI [66.8, 81.8]). 96/139 have |ΔT| > 50 ms.
The denominator is units with ΔT defined — not all 187 resolved units. Dividing by all resolved units counts an undefined ΔT as "not positive" and understates the fraction; that error was found and corrected during review.
This proportion is not a population claim. Units within a session are not independent, and the pooled percentage is weighted by unit count — the two heaviest sessions dominate it. The session-level test is the one that carries inference:
n = 11 eligible sessions, median +20 ms, exact exhaustive sign-flip permutation p = 0.053 (1,024 assignments enumerated).
not significant / insufficient session-level evidence This is not an established null. It is a shift that the available number of sessions cannot establish. See Session-level statistics.