Grid RMS errors exceed endpoint deficits in 24 checks on 12 mouse traces

A new calculation checks a prior result on the same 12 public mouse neuron traces from the Allen Cell Types Database. It uses pairwise covariance of radial threshold indicators and 80-digit Decimal arithmetic, without using the target code or sampling offset bins. All 48 RMS values, 36 continuous area values, 48 zero-offset grid values, and 48 mean identities agree with the target within the planned absolute tolerance of 1e-60 count·µm. The largest RMS difference is 5.8e-77 count·µm. Grid RMS exceeds the endpoint deficit in all 24 trace-spacing pairs at 1 µm and 10 µm spacing. Endpoint RMS is smaller than reference RMS in 18 of 24 pairs, which does not establish correct endpoint geometry. The 12 traces contain 28,827 selected segments, 76 soma connections, and 78 internal radial minima. These numerical checks use the same source observations as the target and do not establish a population error rate or biological causality. Center choice, missing sections, coordinate correction, segment selection, and finite radial windows can change the result. A new sample and specified finite radial windows remain future checks.

Sholl analysis · neuron morphology · grid integration · root mean square error · pairwise covariance · numerical reproduction · Allen Cell Types Database · mouse neuron traces