A factor 1.10 error bound certifies 18 of 69 fitted rise-time reversals in selected Allen cortical inputs

We compared current and voltage response rise-time orders for two inputs of the same sign to one target in Allen cortical slice data. The primary sample had 250 comparisons from 389 measured connections. There were 69 fitted order reversals. An exact error-bound test showed that only 18 must remain if each fitted time can vary independently by a factor from 1/1.10 through 1.10. Three must remain at factor 1.25 and none at factor 1.50. With equal target weights, fitted reversal shares were 21.1% for mouse excitation, 32.5% for mouse inhibition, and 12.5% for human excitation. Their nominal 98.75% donor bootstrap intervals were 10.8 to 32.4%, 20.6 to 45.7%, and 1.7 to 25.4%. Human inhibition had only two inferred donors. The error factors are assumptions, not measured uncertainty. Missing fits, separate recording modes, selected slice circuits and inferred donor IDs prevent a biological reversal-rate claim. Code, scalar data, fit checks, uncertainty draws and a figure support reproduction.

Allen Institute · synaptic rise time · convergent inputs · order uncertainty · current clamp · voltage clamp