Missing-value bounds preserve a positive variability-threshold contrast in four Allen human recordings before distance matching

We tested whether missing resting variability values can reverse a matched reciprocal-minus-one-way threshold contrast in Allen slice data. We used a binary outcome defined by a noise-corrected response spread to median ratio below one. We kept matched groups before outcome filtering and computed exact bounds under all binary assignments of missing values. In 23 human temporal-cortex edges from four recordings and four inferred donors, 21 outcomes were observed. The contrast ranged from 0.100 to 0.308. Distance matching changed the range to -0.025 to 0.183. In 79 mouse visual-cortex edges from 15 inferred donors, 54 outcomes were observed and the range was -0.394 to 0.197. Thus missing values alone cannot reverse the primary human sample sign, but distance matching and a higher threshold remove this constraint. These are finite-sample completion bounds, not confidence intervals or population effects. Observed measurement error, missing-outcome meaning, circuit selection, reverse detection, and small samples limit interpretation. Code, exact assignments, source hashes, a figure, and an independent checker are supplied.

Allen Institute · reciprocal connectivity · missing data · partial identification · synaptic variability